Power Op Amp / Buffer for Low-Ω and Capacitive Loads
← Back to:Operational Amplifiers (Op Amps)
Power op amps and buffers succeed when the real load is treated as R/L/C + cable, and the design is validated against stability, thermal limits, and protection behavior. This page provides practical models, compensation recipes, and repeatable tests so low-Ω, capacitive, and inductive actuators can be driven reliably in the field.
What this page solves (and what it does not)
This page focuses on using a power op amp / buffer as a linear power stage to drive difficult loads (low-Ω, inductive, capacitive, and cable-connected) without surprises: stable control, predictable protection behavior, and safe thermal operation.
Typical loads covered
- Low-Ω resistive loads (high current, headroom and dissipation dominate)
- Inductive actuators (back-EMF, clamp paths, and recovery dominate)
- Capacitive loads and bulk capacitors (phase lag and ringing dominate)
- Long cables (added capacitance + EMI antenna + hot-plug stress)
- Remote parallel capacitance at the load (common trigger for “only sometimes” oscillation)
Common failure modes this page helps debug
- Oscillation / ringing: C-load, cable capacitance, or insufficient phase margin under real load conditions
- Overheating: linear dissipation peaks at certain output voltages and currents (not only at “max output”)
- Current-limit pulsing: protection state interacts with the control loop, causing bursty output
- Slow rise / slow settling: output current, slew behavior, limit modes, and compensation all constrain transients
- Slow recovery after faults: thermal time constants + protection sequencing dominate restart time
- EMI sensitivity: loop area, cable routing, and clamp placement change both stability and emissions
Scope boundary (keeps this page focused)
- High-voltage op amp SOA deep-dive (±15 V and above): only referenced here; detailed SOA is handled in the High-Voltage Op Amp page.
- Howland / 4-quadrant V→I sources: only the concept of current drive is introduced; full transconductance design stays on the Howland Source page.
- ADC/FDA low-distortion driving: distortion/matching topics belong to ADC Driver / FDA pages, not this actuator-load page.
- Full loop-stability theory derivation: this page uses practical “load-first” recipes; theoretical derivations belong to the Stability & Compensation page.
The same device can look “fine” on the bench and fail in the field because the real load model (cable + C + back-EMF) changes the loop phase and triggers protection states. The rest of this page treats stability, protection, thermal, and layout as one coupled system.
Where power op amps sit in a system
A power op amp / buffer is best treated as a linear power stage inside a closed loop. Most failures are not “mysterious silicon behaviour” but predictable outcomes of: loop gain, load impedance vs frequency, and protection/thermal state changes.
System blocks to think in (left-to-right)
- Setpoint source: MCU/DAC (or filtered PWM) defining the target value
- Error amp / loop filter: sets bandwidth, phase margin, and transient shape
- Power op amp / buffer: delivers current and swing under headroom and thermal limits
- Load: R/L/C/cable effects dominate phase and stress
- Sense path: voltage or current sensing (Kelvin routing often determines accuracy and stability)
- Protection / clamps: current limit, thermal foldback, and output clamps can change the effective plant
Two operating modes (concept only)
- Voltage drive: the loop regulates Vload; current is set by the load (watch dissipation and current-limit behaviour).
- Current drive: the loop regulates Iload; compliance voltage, clamp paths, and inductive energy handling become critical (full transconductance designs live on the Howland page).
Single-supply vs dual-supply pitfalls (the ones that cause field returns)
- “Can it really hit 0 V under load?” Output headroom often worsens at high current.
- “What happens during back-drive?” Inductive loads and external sources can pull the output beyond rails.
- “Does protection alter the loop?” Limit/foldback modes can create pulsing unless the loop and clamps are designed together.
When protection or thermal limiting activates, the power stage becomes a different plant than the one assumed in small-signal analysis. Designs that remain stable and predictable treat sense routing, clamp paths, and limit modes as part of the loop—especially with cables, bulk capacitance, and inductive energy.
Load models that matter (R, L, C, cable) — and why they break stability
Real loads are not “a resistor.” Their impedance changes with frequency, wiring, and protection paths. Stability issues appear when the load adds phase lag, changes the effective plant, and pushes the loop into oscillation or limit cycling.
Model → symptom → what to check (practical, load-first)
Low-Ω resistive load (R)
- What it really means: high current demand makes dissipation and headroom the dominant limits.
- Typical symptom: “works at light load” but clips, droops, or enters limit modes at the target current.
- Check first: peak vs continuous current, source/sink asymmetry, and supply droop at load steps.
Inductive actuator (R + L)
- What it really means: current changes create back-EMF; the output can be forced beyond rails during transients.
- Typical symptom: unexpected resets, protection trips, or damage when the load is unplugged or rapidly driven.
- Check first: freewheel/clamp path (where energy returns), and whether the return loop is short and controlled.
Capacitive load (R + C)
- What it really means: added poles reduce phase margin; inrush current can trigger limit behaviour.
- Typical symptom: ringing or oscillation that depends on capacitor value, cable length, or probe connection.
- Check first: isolation resistor (Riso) and snubber placement close to the output/load interface.
Long cable (R + C + EMI)
- What it really means: cable capacitance behaves like extra C-load; the cable also couples noise and radiates.
- Typical symptom: stable on the bench, unstable in the field; hot-plug and routing change the behaviour.
- Check first: loop area, connector-side protection, and damping strategy (driver-side vs connector-side).
Why stability breaks in real life (without heavy theory)
- Extra phase lag: C-load and cable capacitance add poles that reduce phase margin at the gain crossover.
- State changes: current limit or thermal foldback turns the power stage into a different plant than assumed.
- Uncontrolled return paths: clamp energy and EMI return loops can inject error into the sense path and loop filter.
The next section turns these load models into concrete drive limits: current, headroom, and recovery under overload and limit modes.
Output stage & drive limits: current, headroom, swing, and recovery
“Drive capability” is set by output current, headroom under load, and what happens during and after overload. A design can meet the DC target yet still fail because limit modes and recovery time reshape the output behaviour.
Practical limits (each one has a measurable symptom)
1) Output current: peak vs continuous; source vs sink
- Why it matters: peak current may look strong, but continuous current and thermal limits define real capability.
- Typical symptom: one edge is slower (charge vs discharge), or one polarity clips first (source ≠ sink).
- Quick check: step the setpoint with real load and record both rising and falling behaviour.
2) Headroom and swing under load
- Why it matters: rail-to-rail claims do not guarantee rail-to-rail swing at high current.
- Typical symptom: output “stops short” of the target near a rail even with seemingly adequate supply.
- Quick check: sweep output setpoint near both rails at the target load current and measure the error.
3) Overload recovery: saturation, limit, and thermal foldback
- Why it matters: leaving linear operation can change loop gain and the effective plant for seconds.
- Typical symptom: a short-circuit is removed but the output remains slow, distorted, or pulsing.
- Quick check: intentionally hit current limit once and measure time-to-recover to within ±x% of the target.
4) Capacitive transients: inrush peaks and “mystery” clipping
- Why it matters: charging a large C-load forces an inrush current spike that can trigger limit modes.
- Typical symptom: the output rises in steps or flattens during the edge (limit-mode shaping).
- Quick check: repeat the same step with different capacitance and cable length to see sensitivity.
If the operating point approaches the limit boundary, the output stage may enter current limiting or thermal foldback, reshaping the waveform and the loop behaviour. Robust designs validate both polarities (source/sink), near-rail headroom, and time-to-recover after a forced overload event.
Thermal reality: power dissipation, RθJA, heatsinking, and derating
For linear power stages, “missing power becomes heat”. The maximum continuous drive is therefore set by a thermal budget: power loss → thermal resistance → temperature rise → junction margin.
A simple, reusable thermal budget (linear drive)
- Step 1 — Estimate loss: Ploss ≈ (Vsupply − Vload) · Iout (the drop across the power stage becomes heat).
- Step 2 — Pick a thermal path: choose θ values that match the mechanical reality (package-to-board, board-to-air, heatsink-to-air).
- Step 3 — Convert to temperature rise: ΔT ≈ Ploss · θ and Tj ≈ Tamb + ΔT.
- Step 4 — Keep margin: ensure a safe gap to the junction limit under worst-case ambient and load.
Thermal problems often appear at operating points with large voltage drop and high current (not only at the highest output voltage). A stable waveform on the bench can become limit-cycled in an enclosure because the thermal budget silently collapses.
θJA vs θJC (what they mean in practice)
- θJC: junction-to-case (or junction-to-exposed pad). Useful when heat is pulled into a defined mechanical path (heatsink / metal chassis).
- θJA: junction-to-ambient for a specific test setup. Strongly affected by PCB copper, via arrays, airflow, and enclosure constraints.
- Engineering rule: θ values are not universal constants; treat them as a starting point and validate on the real board.
Continuous vs pulse (why peak ratings can mislead)
- Continuous drive: governed by steady-state heating (θJA/θJC dominate).
- Pulsed drive: short pulses can exceed steady-state limits if pulse width and duty cycle keep the average loss within margin.
- Actionable takeaway: peak current numbers are only meaningful when paired with pulse conditions (width / repetition / duty).
A reliable design treats thermal as a first-class requirement: if the thermal budget fails, protection states activate and the output waveform changes. The next section explains why “protection” can look like oscillation, pulsing, or slow recovery.
SOA and protection: current limit, short-circuit, thermal shutdown, clamp paths
Protection is not a simple “safety switch.” It behaves like a state machine that can reshape the plant and the waveform. When a limit mode turns on, loop gain and phase effectively change, which can create pulsing, jittery output, or slow recovery.
Common current-limit behaviours (what they look like at the output)
Constant current limit
- Output symptom: edges flatten; the load voltage becomes “whatever the limited current can support.”
- Design implication: large C-load steps may instantly hit the limit and reshape transient behaviour.
Foldback current limit
- Output symptom: “hiccup” or repeated attempts to rise; some loads never start cleanly.
- Design implication: the system may oscillate between “almost starts” and “backs off,” especially with cables and capacitance.
Thermal foldback / thermal limiting
- Output symptom: performance degrades with time; the same command produces less current as the device heats.
- Design implication: field failures often look random, but correlate with enclosure temperature and airflow.
Short-circuit and reverse energy (inductive back-drive)
- Short-circuit: the device may enter current limit immediately, then heat up and transition into thermal limiting or shutdown.
- Reverse energy: inductive loads or external sources can push the output above/below the rails; clamp paths determine where that energy goes.
- Clamp paths matter: returning energy into a long or shared ground can inject noise into sensing and destabilize the loop.
What to validate on the bench (to avoid field surprises)
- Trip points: current limit threshold and how it changes with temperature.
- Waveform in limit: flat-top, pulsing, or burst behaviour under the real load model (cable + C).
- Recovery time: time-to-return within ±x% of the target after a forced short or overload.
- Back-drive event: unplug / fast reversal / inductive kick response and clamp heating.
A robust design assumes protection will activate at least once. The goal is not to avoid every trip, but to ensure the system remains predictable: clean limit behaviour, controlled clamp paths, and a measured recovery time that matches the application.
Stability for real loads: practical compensation recipes (without heavy theory)
Real actuator and cable loads add phase lag, inject energy, and change the effective plant. This section focuses on hands-on fixes organized as symptom → action → tradeoff, so stability can be recovered without heavy theory.
C-load (capacitance on the output): the most common trigger
Symptom
- Stable at light load, but rings or oscillates when a capacitor, long cable, or “mystery” probe capacitance is present.
- Waveform changes noticeably with different cable lengths, connector routing, or probe ground lead length.
Action (try-first order)
- Add an isolation resistor (Riso) at the driver output: start small and increase until ringing clearly decays. Place Riso close to the op amp output to isolate the external capacitance from the output stage.
- If ringing remains narrowband, add an RC snubber: use it as high-frequency damping to absorb the oscillation energy. Place it where the loop is shortest (typically near the driver output / interface node).
- Only then consider slowing the loop: bandwidth reduction or gain changes are effective but can degrade dynamic accuracy and response time.
Tradeoffs
- Riso: improves phase margin, but increases output impedance (extra droop under load and more dissipation at high current).
- RC snubber: adds high-frequency loss and heat, but can tame stubborn narrowband ringing and EMI peaks.
Inductive loads (back-EMF): stability and protection interact
- Symptom: spikes, pulsing, or resets during turn-off, reversal, unplug, or rapid setpoint steps.
- Action: define a controlled freewheel/clamp path (where the energy returns) and keep the clamp loop short.
- Tradeoff: clamping protects silicon but can inject energy into rails or ground if the return path is not controlled.
Long cables (capacitance + antenna): damping and loop area
- Symptom: stable on a short bench lead; unstable or noisy with the real harness in the field.
- Action: add damping and make the output-return loop small. Choose driver-side damping for direct stability benefit; choose connector-side damping for better entry-path control.
- Tradeoff: heavier damping reduces EMI and ringing but slows edges and increases power loss.
Gain setting: why lower closed-loop gain is often harder
- Practical rule: many amplifiers are most sensitive near low closed-loop gain (often near unity), where load-induced phase lag can consume the remaining margin.
- Action: if the system allows it, use a higher closed-loop gain or select a device specified as stable for the intended gain range.
- Tradeoff: raising gain changes the system scaling and may require recalibration of the control chain.
Try-first checklist (fast convergence)
- Stabilize C-load with Riso at the driver output.
- Use a targeted RC snubber if ringing remains narrowband.
- Define and tighten clamp/freewheel return paths for inductive loads.
- Then adjust gain/bandwidth only if needed (accepting response-time tradeoffs).
These fixes work best when placed intentionally. Driver-side compensation improves loop margin directly, while clamp/freewheel design must keep high di/dt return loops short and predictable.
Output protection & load interface: diodes, TVS, back-EMF, and cable EMI
Output failures in the field are often caused by where protection is placed and how clamp current returns, not by missing parts. Cable entry paths, inductive kick, and back-drive events demand a controlled interface design.
Inductive load freewheel paths: return-to-rail vs return-to-ground
- Return to supply rail: can reduce ground noise, but may lift the rail and disturb other circuits if the rail impedance is high.
- Return to ground: avoids rail lift, but can inject large current spikes into ground, corrupting sensing and control references.
- Practical goal: keep the freewheel/clamp loop short and ensure the return path does not cross sensitive analog ground.
TVS and clamp choices: clamp-to-which-rail is a system decision
- Clamp destination matters: clamping to a rail means injecting surge energy into that rail. The rail decoupling and routing must handle it.
- Return path matters more than the symbol: long clamp loops create voltage spikes, EMI, and false triggers.
- Placement priority: protect near the entry (connector/cable) when the threat originates from the harness.
Back-drive and reverse output voltage (what causes it)
- External sources: another driver, shared bus, or powered load can force the output beyond the intended range.
- Inductive return energy: the load itself can push the output above/below rails during rapid current changes.
- Interface goal: ensure the output pin never sees uncontrolled reverse stress by defining clamp paths and loop geometry.
Placement rules (avoid mixing “entry protection” with “loop compensation”)
- Protect near entry: TVS/ESD parts work best when placed close to the connector that introduces the threat.
- Compensate near driver: Riso/snubber are part of the loop behaviour and are most effective close to the op amp output.
- Keep clamp loops short: high di/dt loops should not run across sensitive ground or sense nodes.
Entry protection belongs near the connector, while loop-shaping parts (Riso/snubber) belong near the driver output. Keeping high di/dt clamp loops short prevents both EMI and false stability failures that appear only with real harnesses.
Layout & grounding for high current + stability
High-current layout is not only about efficiency. It sets loop area, ground bounce, clamp return behaviour, and the stability margin under real loads. This section focuses on principles; a checklist-style review appears later.
Principle 1 — Minimize the high-current loop area
- Think in loops: OUT → load → RETURN is the loop that radiates and bounces the reference.
- Route as a pair: keep the return path adjacent and continuous; avoid detours and slot crossings.
- Keep sensitive nodes out of the loop: do not let load return currents flow under sense/feedback networks.
Principle 2 — Separate power return from analog reference (Kelvin / star logic)
- Kelvin sense: feedback and current-sense signals should connect at the true measurement points, not on a shared power ground.
- Star intent: define a reference point where small-signal ground meets power return, so load current does not modulate the reference.
- Common symptom: output looks “unstable” only when load current changes; the real cause is ground injection into the sense path.
Principle 3 — Decoupling priority (bulk + HF ceramic)
- HF ceramic first: place small ceramics closest to the output-stage supply pins to supply fast current with low inductance.
- Bulk supports energy: larger capacitors maintain rail stability during longer transients; keep their loop short and return path clear.
- Stability tie-in: long supply loops increase effective rail impedance and can worsen ringing and limit-cycling.
Principle 4 — Thermal-aware copper (keep the thermal budget real)
- Spread heat: use copper area and via arrays under/around the power package to reduce local hotspots.
- Plan the neighborhood: keep drift-sensitive networks away from the hottest copper and clamp return loops.
- Match the enclosure: layout decisions should reflect real airflow and mounting, otherwise θ assumptions break.
Common failure patterns (layout-driven)
- Ringing changes with probe grounding or cable routing.
- Readings shift with load current even when the command is constant.
- Protection pulsing appears only in enclosure conditions.
- Clamp/TVS “works” but EMI is worse due to long return loops.
A stable power stage is often defeated by layout: uncontrolled return paths reduce effective phase margin and inject load current into the reference. Treat the output-return loop and the sense reference as first-class design objects.
Selection flow: from load requirements to part choice
Part selection is easiest when requirements are treated as constraints. Start with the real load and environment, translate them into hard limits (current, headroom, stability, thermal, protection), then choose the right class of solution.
Step 1 — Capture the inputs (the “must-fill” fields)
- Load model: R/L/C, maximum capacitance, cable length, and any back-EMF behaviour.
- Current profile: peak vs continuous, duty cycle, startup/stall conditions.
- Supply: single/dual rails, headroom margin, and whether reverse energy into the rail is acceptable.
- Dynamics: required response speed (slew/bandwidth as means, not goals).
- Stability constraints: allowable C-load and whether Riso/snubber tradeoffs are acceptable.
- Thermal + protection: θ assumptions, enclosure temperature, short behaviour, thermal shutdown, and reverse output tolerance.
Step 2 — Convert inputs into constraints (pass/fail gates)
Iout (continuous)
Must hold at worst-case Tamb without thermal foldback.
Headroom
Output swing under load must meet range targets.
C-load stable
Must remain stable with cable + capacitance (or accept Riso/snubber).
SOA / protection
Limit behaviour and recovery must be acceptable.
Back-drive tolerance
Reverse energy must not overstress the output pin or rails.
Step 3 — Choose the right bucket (avoid forcing the wrong class)
Power op amp
Best when a linear, closed-loop solution must hold accuracy and predictable protection behaviour.
Buffer / driver
Useful when drive strength dominates and precision requirements are modest.
External stage
When current/voltage/thermal limits exceed a reasonable linear IC, move power outboard and keep the op amp as a controller.
A strong candidate is the one that passes the constraints with margin: continuous current at temperature, stable behaviour with real capacitance and cable, and a protection/recovery profile that matches the application.
Engineering checklist & validation tests (bring-up to production)
This section provides a repeatable bring-up ladder and validation tests that turn real-load stability, protection behaviour, and thermal limits into measurable pass/fail criteria. The goal is to reduce surprises in enclosure, on cable, and under temperature.
A) Bring-up ladder (progressive load steps)
The sequence below prevents “jumping straight to the real harness” and makes failures diagnosable. Each step includes the minimum observation points and a simple metric to record.
Step 0 — No load
- Stimulus: small step and slow ramp in command (avoid immediate rail hits).
- Probe points: Vout, Vsupply at the IC pins, feedback/sense node.
- Pass metric: no sustained ringing; output follows command without “hunting”.
- Common fail: ringing changes with probe grounding (measurement loop issue) or idle oscillation (loop margin too low).
Step 1 — R load (pure resistive)
- Stimulus: step response at a few amplitudes; include a near-worst-case current step.
- Probe points: Vout at load, Vout at IC pin (if accessible), Vsupply droop at IC pins.
- Pass metric: overshoot (%) and settling time to a defined band (e.g., ±1%).
- Common fail: headroom collapse under load current (swing shrinks) or supply loop/decoupling causes ringing.
Step 2 — C load (capacitance / cable-equivalent)
- Stimulus: square wave or step into the intended operating range.
- Probe points: Vout at the output node, Vout at the capacitor/load node, feedback node.
- Pass metric: ringing cycles until decay to a threshold (e.g., <10% of initial) and peak-to-peak ringing.
- Common fail: narrowband sustained ringing; fix order typically starts with Riso, then targeted snubber.
Step 3 — Real load (actuator / inductive or mixed)
- Stimulus: realistic command transitions (start/stop, polarity reversal if applicable, worst-case step).
- Probe points: Vout, clamp node (TVS/diode node), Vsupply, sense/feedback node.
- Pass metric: clamp peak voltage and recovery time after large transients.
- Common fail: protection pulsing (current limit/foldback) mistaken as “oscillation” — confirm by correlating current and temperature.
Step 4 — Cable / EMI (real harness and entry paths)
- Stimulus: repeat Step 2/3 waveforms with short/medium/long cable lengths.
- Probe points: Vout at driver side and at connector/load side, TVS current/return loop (by observation of spike shapes).
- Pass metric: stable response across cable variants; no unexplained resets; ringing remains bounded.
- Common fail: entry protection placed correctly but return loop too long, worsening spikes and EMI.
B) Stability validation (make “stable” measurable)
- Step response: record overshoot (%), peak-to-peak ringing, and settling time to a chosen band.
- Cable sweep: validate at multiple cable lengths; document routing and shield termination condition.
- Temperature sweep: cold start and hot steady-state can shift margin; re-check ringing and recovery.
- Supply disturbance: apply controlled rail droop or ripple (within safe bounds) and confirm no limit-cycling.
- Measurement hygiene: use short ground springs; a long probe ground can create “phantom ringing”.
Recommended record fields
overshoot(%), ringing_pp(V), ringing_cycles, settling_time(ms), cable_length(m), Tamb(°C), Vsupply(V), note(tag).
C) Thermal & protection validation (separate “loop issues” from “state-machine behaviour”)
- Steady-state temperature rise: record ΔT at a defined load current and ambient condition; note time to steady state.
- Current limit point: measure the limiting threshold and its drift with temperature.
- Short-circuit recovery: time from short removal to linear output recovery; watch for pulsing or latch-like behaviour.
- Unplug / reconnect: observe spikes at the clamp node and confirm recovery without resets or output latch-up.
Production-friendly metrics (examples)
- ΔT(°C) at Icont, Tamb; hotspot or case measurement method documented.
- Ilimit(A) at Tamb low/nom/high.
- short_recovery_time(ms) and post-event offset shift.
- max_clamp_peak(V) during unplug/reconnect or worst command step.
D) Validation fixture BOM (example parts with orderable part numbers)
These are common, widely available examples for building repeatable loads and protection fixtures. Equivalent parts are acceptable; the fixture goal is controlled R/L/C, short clamp loops, and documented cable conditions.
Resistive loads (R load)
- Ohmite HS50 series (aluminum-housed power resistors; choose value/wattage): example family “HS50”.
- Vishay LPS series (power resistor family; choose value/wattage): example family “LPS”.
- Bourns PWR221T series (power resistor family): example “PWR221T-20R0F” (value can be swapped).
Capacitive loads (C load / cable-equivalent)
- WIMA MKP2 series (film capacitor family; stable “clean C” for ringing tests): example family “MKP2”.
- KEMET R82 series (film capacitor family): example family “R82”.
- Murata GRM series (MLCC family; note DC bias effects): example family “GRM”.
Compensation parts (Riso / snubber)
- Vishay CRCW1206 series (thick-film resistors for Riso/snubber-R): example family “CRCW1206”.
- Murata GRM31 series (MLCC for snubber-C in a robust size): example family “GRM31”.
- TDK C3216 series (MLCC alternative family): example family “C3216”.
Clamp / protection parts (for interface and back-EMF tests)
- Vishay SS54 (Schottky diode example; select voltage/current to match the load).
- onsemi MBR series (Schottky family; choose rating as needed): example family “MBR”.
- Littelfuse SMBJ series (TVS family; select standoff/clamp voltage): example family “SMBJ”.
- Littelfuse SMAJ series (TVS family alternative): example family “SMAJ”.
- Nexperia PESD series (ESD/fast transient protection family near connector): example family “PESD”.
Cable variants (for Step 4 repeatability)
- Length set: 1 m / 3 m / 10 m (or closest practical trio) to expose cable-capacitance sensitivity.
- Condition tags: shielded vs unshielded; shield termination method recorded.
Recording consistent metrics at each ladder step makes design changes traceable. A stable design should remain stable across cable variants, temperature, and supply disturbances, without protection pulsing or slow recovery.
Request a Quote
FAQs – Power Op Amp / Buffer
Short, practical answers that close common long-tail questions without expanding the main content boundary. Each item includes a concise answer plus three key checks and three fixes.
