Versionshinweise

1.5.0

FFB-Bridge 1.5.0 makes the vibration level of a force-feedback base an independent control, fixes three bugs that discarded saved hardware limits on launch, moves the MOZA AB9's fixed-wing trim back to the standard centering spring, softens direction changes and end-of-travel braking, makes Flight Check park cleanly at the ends of travel on direct-drive bases, links the Dashboard's active-effect chips to their Tuning sliders, adds mouse back and forward navigation, and lets beta-channel installs move to a newer stable release.

Vibration and detail is its own control

• Hardware, Vibration & detail sets the strength of engine rumble, buffets, runway rumble, shudders, and transient cues directly. Since 1.4.1 these cues were limited to the lower of the pitch and roll output ceilings, so a base run at a conservative 35% ceiling for control authority also had its vibrations held at 35%. On a MOZA AY210 flying the C172 profile that left cruise engine rumble below the level the yoke can render. The slider is now an absolute ceiling from 0 to 100% of device command, independent of the axis ceilings. Left untouched it follows the lower axis ceiling exactly as before. Values above the recommendation need the same acknowledgement as the axis sliders. Confirmed on a physical AY210.
• Older vibration settings carry over. The multiplier-style value from 1.4.0 is converted to the equivalent absolute ceiling on first launch.

Saved hardware limits apply on every launch

• Per-device output limits and the vibration ceiling are applied when the device opens. On macOS, and on any launch where the device could not be identified before opening, saved limits were silently replaced by the automatic envelope, so a yoke saved at 99% pitch ran at 35% until a slider was touched. This was the cause of a MOZA AY210 feeling weak on 1.4.4 on the Mac.
• The Hardware sliders no longer shrink a saved value. The slider range initialized to the recommendation before the device had opened, clamped the restored value, and saved the clamped number back.
• The above-recommendation acknowledgement is remembered per device. It was previously inferred from live state, which raced the device open and made every slider touch clamp and re-save the value.

MOZA AB9: fixed-wing trim on the standard centering spring

• Trim no longer removes the Forces-section feel on the AB9. Since 1.4.1 the AB9 rendered fixed-wing trim through a firmware route that, on this base, suppresses the centering load, airspeed feel, and stick weight whenever the Trim slider is above zero, while end stops and effect gains survive. Four support bundles from a P-51D report showed the exact signature. Trim now shifts the standard centering spring, as it does on every other base. The Safety Stop that could fire when the trim setting crossed zero was a side effect of that firmware boundary and is gone with it. Helicopter force trim on the AB9 is unchanged.

Softer direction changes and end-of-travel braking

• Force reversals complete in about half the time. The reversal smoothing introduced in 1.4.1-rc3 is relaxed from 4 to 8 units of travel per second, so a typical reversal takes about 75 ms instead of 150 ms.
• The end-of-travel brake starts later. It now engages in the last 15% of travel instead of the last 30%. Both changes were flown on a physical AY210.

Flight Check on direct-drive bases

• Pitch and roll direction tests park at the end of travel without bouncing. On the MOZA AB9 and AB6 the held test force reached the stop, chattered, and bounced back. Bench tests now reduce the outward drive progressively over the outer 40% of travel and with outward speed, so the stick slows before the stop and settles. Live flight is not affected. Verified on a MOZA AB6.
• The first Start Test no longer arms and immediately disarms. On MOZA bases the first arm after launch writes and verifies about 34 firmware settings over the serial channel. The control-loop watchdog measured the whole sequence against its 1.5 second deadline and tripped, disarming the base; a second arm was faster and passed. Each write is now timed on its own, so a stalled serial port is still caught while a healthy sequence completes.

Dashboard and navigation

• Active-effect chips open their Tuning slider. Clicking a chip on the Dashboard switches to Tuning, opens Advanced if the slider is gated, expands the section, centres the row, and highlights it until you leave the page. Rotation kick, Surface follow, and Sideslip have no slider and remain display-only. A blue pointer hint below the grid makes this shortcut visible without relying on hover alone.
• Mouse back and forward buttons navigate between pages, the way they do in a browser.
• The DEVICE lamp on the arm gauge reflects a physically connected stick. It previously showed Device Ready with no stick attached.

Updates

• Beta-channel installs see stable releases. A beta client now checks both the beta and stable feeds and offers whichever version is newer, so it is no longer stranded on a prerelease after the stable release supersedes it.

1.4.4

FFB-Bridge 1.4.4 repairs a serious 1.4.2 regression on the MOZA AY210 that left the yoke without centering load, airspeed feel, or stick weight while engine and ground vibrations continued, adds a base damping floor and working base-resistance controls for the AY210 family, fixes helicopter autopilot follow never moving the stick on non-MOZA devices, and adds conditional support for the MOZA AY90.

MOZA AY210: stick feel restored

• Centering load, airspeed feel, and stick weight render again. In 1.4.2 the AY210's static forces were routed through a firmware mechanism that accepts commands but does not physically render them on this base, and the effects created at arm were additionally silenced by the follow setup, so only engine and ground vibrations came through. Trim and the airspeed load now render through the standard spring and direct force channel, the same progression the base had in 1.4.0: resistance starts at low airspeed and builds continuously with speed. Confirmed on a physical AY210.
• Forces survive autopilot engage and disengage. Entering or leaving autopilot follow is a firmware boundary on MOZA bases after which previously loaded effects go silent. Every effect is now refreshed at that boundary, on the AY210, AY90, and AB9. Autopilot follow itself still runs through the firmware servo, which is unchanged.
• Trim behaves like 1.4.0. The elevator trim wheel shifts where the yoke settles through the normal centering spring on cable-linked aircraft.

MOZA AY210 and AY90: base damping while armed

• The base keeps a minimum resting resistance while armed. Arming zeroes a MOZA base's own damping so the bridge can render its own feel; on the AY210 family nothing was put back, and with forces restored an undamped direct-drive yoke could oscillate when released away from centre. While armed the base now holds a minimum of 10% damping, 5% friction, and 5% inertia, measured on a physical AY210.
• The Base resistance sliders now work on these bases. Hardware, MOZA, Base resistance saved values on the AY210 family previously had no physical effect. Your setting now replaces the default minimum; an aircraft profile can ask for more but never less.

Helicopter autopilot follow moves the stick

• Autopilot follow works with force trim configured. On helicopter profiles the force trim system owned the centering spring outright, so the autopilot follow cue never reached the device on the SideWinder FFB2 and other non-MOZA sticks, in both simulators, even with the follow slider raised. While the autopilot is engaged it now owns the spring; holding the trim release still frees the cyclic; and on disengage the force trim holds the cyclic where the autopilot left it, the way a coupled autopilot flies through the trim system.

New device: conditional support for the MOZA AY90

• The MOZA AY90 FFB base (346E:1003) is supported conditionally. It runs as an AY210-family base: raw HID and DirectInput on Windows, evdev on Linux, IOHID on macOS, firmware configuration ownership, autopilot follow, and the same damping floor. Support is based on three approved ffb-probe reports and the AY210's validated behavior; it has not yet been validated on a physical AY90. If you fly one, reports of anything that feels or behaves wrong are welcome through Support.

1.4.2

FFB-Bridge 1.4.2 is the stable roll-up since 1.4.1. It adds driverless support for the Logitech G940 on all three operating systems, completes MOZA AB6 autopilot follow through the official firmware path, stops opening the app from changing how a MOZA base feels before you arm, keeps the AB6 cyclic in your hand in helicopter mode, repairs Linux faults that could leave a stick unusable after the first disarm, stops forces being produced from simulator states that are not flight, grounds several aircraft cues in telemetry measured from the simulators rather than inferred, and fixes two long-standing feel defects: a phantom turbulence burst after pitch inputs and spoiler buffet that went silent at low airspeed.

Logitech G940

• The G940 is driven directly, without Logitech's driver. FFB-Bridge now speaks the stick's own force protocol on Windows, Linux, and macOS: 16-bit constant force, the native spring, the damping channel, and native centering. On Windows this removes the need for Logitech Gaming Software 5.10, whose 2010 kernel drivers current Windows Memory Integrity blocks. On Linux the native spring and damping channel are available for the first time, because the kernel driver exposes neither. On macOS the stick has force feedback at all for the first time. Tested on all three operating systems.
• The platform driver paths remain as automatic fallback. When the vendor path cannot be opened, FFB-Bridge falls back to the driver path and records why in the session log. On Linux the vendor path needs a writable hidraw node, so the generated udev rule has changed; Support → Health checks detects the older rule and offers to replace it.
• Rumble, buffets, and mechanical cues are synthesised at 200 Hz and folded into the constant-force output, because this hardware renders no periodic effects of its own. Logitech's driver does the same at roughly 62 Hz.
• Friction and Inertia are not offered on this stick. Damper, Friction, and Inertia all write the same field in its protocol, so only Damper is exposed.
• Axis range is taken from the stick's HID descriptor. The Linux kernel advertises a range the device does not use, which made pitch and roll readings wrong.

Brunner

• Two more grips are recognised. The FFB-G with a MOZA MH16 grip (25BB:0120) and the CLS MZ in DirectX mode with an MH16 grip (25BB:0121) are both handled as the FFB-G base. Brunner allocate a separate product id per grip, so a base wearing a grip that is not listed is not recognised at all; report the id and it can be added.

Forces stop when the simulator is not flying

• The MSFS 2024 main menu no longer produces forces. The menu publishes an incoherent frame — low indicated airspeed alongside a high Mach number and a true overspeed warning — which an armed stick rendered faithfully. The menu camera states were numbered for MSFS 2020 and did not match 2024, so the existing guard never fired.
• Time compression above 2x leaves centering only. Rotation rates, accelerations, and the turbulence window all arrive already multiplied by the simulation rate, so at high compression every rate-derived force ran far too fast. Forces stop immediately on the change and return through the normal arm ramp.
• X-Plane is only reported connected when X-Plane answers. Selecting X-Plane while it was closed could show a connection, arm the stick, and compute forces from an empty snapshot, cycling between armed and disarmed every few seconds. FFB-Bridge now waits for a real response before reporting a connection and publishes nothing until the first genuine packet arrives.
• Auto-arm requires a real force-feedback device. If the selected device is missing, the status line, summary, and compact chip say so instead of arming against the in-memory stand-in.
• The stall buffet is silent on the ground. A parked aircraft with the brakes set could buffet continuously; the stall buffet was the only aero cue with neither a ground gate nor an airspeed floor.

Aircraft feel

• Mach buffet is graded against the aircraft's own limit. The buffet now builds against the barber-pole Mach the aircraft publishes, in the same 94–102% window the overspeed cue uses in the airspeed domain. A profile's fixed Mach window remains as the fallback.
• The stick-shaker airspeed floor comes from the aircraft's design stall speed instead of a fixed 40 knots, which was unusable for a Cub, a glider, or a Tiger Moth. Aircraft that publish no usable value keep the old floor.
• The touchdown thump uses the impact speed the simulator latches at contact, so it is correct on sloped ground and survives simulators that zero vertical speed on the contact frame.
• Vortex ring state uses the simulator's own indication alongside the existing sink/collective/rotor heuristic, whichever is stronger. Confirmed on a Bell 407.
• Gear transit has a cue of its own. The three seconds of legs swinging between the lever and the airflow over extended gear produced nothing before. It follows the gear motor circuit where the simulator reports one and actual gear movement everywhere else, with a new Gear travel control that silences it at zero.
• Flap travel combines actuator and movement evidence. The flap motor circuit and actual surface movement are now treated as independent evidence for the same event rather than one overriding the other, which restores the travel cue on aircraft whose circuit under-reports.
• Turboprop propwash follows engine torque, not propeller RPM. A governed free-turbine prop idles near 60% and holds near 100% in flight at any power, so the RPM proxy produced a strong aft pull from a parked Grand Caravan. Piston aircraft keep the RPM drive unchanged.
• Hydraulic-system loading. When an aircraft reports hydraulic integrity below 100%, control loads and centering rise in proportion, through a new Hydraulic loss loading control. It is off by default, exempt on fly-by-wire aircraft, and enabled on the H125 starter.
• A pitch input no longer triggers a phantom turbulence burst. The turbulence cue measured the spread of recent G readings, so the smooth G change of the pilot's own maneuver registered as turbulence and produced a short shake right at level-off; on strong direct-drive bases that shake could rock the pitch axis for a few cycles. Turbulence now measures only the variation around the current G trend, so a deliberate maneuver is silent while genuine choppy air feels exactly as before at the same setting.
• Spoiler buffet is continuous while the surfaces are out. It begins at the first crack of deployment instead of after a dead zone, no longer falls silent at low airspeed (deployed glider airbrakes at approach speed used to compute to almost nothing, leaving only a cue while the boards moved), and grows with both deployment and airspeed to the configured strength. Deployment movement itself also renders a travel cue.
• Trim moves the stick only on aircraft whose controls are cable-linked to the trim system. Airliners trim the stabilizer, not the control column: a real 737 or A320 column stays neutral as trim runs, so those profiles now keep a neutral-centered spring while trim force relief continues to work on every type.
• Pitch/roll force balance applies on MOZA follow bases. The per-axis elevator/aileron force scales acted only on the directed force channel, which those bases do not use for centering, so a profile's roll-heavy balance rendered as equal. The balance now carries onto the device-owned spring.
• The gear thump fires on retraction as well as extension.

Autopilot follow

• MOZA AB6 AP Follow runs on the official firmware path and passed live flight acceptance. FFB-Bridge takes durable, read-back-verified ownership of the base's complete autopilot tuning family (speed, ratio, breakout force, and transition range) and restores the exact original values on disengage, shutdown, and crash recovery. Flight testing on a C172 corrected the conditioning three times before acceptance: the breakout force is set to 25% (the factory 50% produced a hard centre detent), the follow spring is doubled for the 6 N·m base (the 12 N·m AB9's strength could not hold the stick against gravity), and the spring engages before the rest of the engage sequence (the handoff otherwise left the stick unsupported for most of a second and it fell forward into a dive the simulator read as pilot input). Engage, tracking, grabbing the yoke under AP, disengage handback, and trim feel all passed on the attached base. Fixed-wing trim renders on the normal spring path because the AB6's firmware trim channel stores targets without moving both directions symmetrically. MSFS AP Follow uses a bounded washed flight-director/surface reference; X-Plane retains its full-range total-control-ratio reference.
• X-Plane autopilot follow uses the owned-axis path on ordinary aircraft again. This X-Plane answers subscriptions for datarefs an aircraft does not have with a constant zero, which made every aircraft look like it drives its own controls; the full-range owned-axis follow path therefore never engaged anywhere and every aircraft silently ran the 20% coupled ceiling. Recognition now requires a real non-zero observation, so ordinary aircraft get the full saved authority range and the Zibo is still detected the moment its own autopilot engages.
• Fixed-gear profiles no longer inherit gear-motor buffet. Gear-motion was added after saved profiles already existed, and its former 18% constructor default silently enabled the cue when an old profile omitted the field. A missing gear-motion member is now silent; retractable built-in starters opt in explicitly, while the C172 and other fixed-gear starters remain at zero.
• Follow renders at the authority it is set to again. A source change on 2026-08-10 left the X-Plane owned-axis path capped at half the saved value. It now renders the full saved range, and the coupled path used by MSFS was raised from 10% to 20% after a SideWinder sweep in which both a C172 and a King Air tracked cleanly at 20% and the C172 hunted at 30%.
• The Zibo 737-800X is supported. Its autopilot is detected — it replaces X-Plane's own and leaves the standard servo flag off, so follow behaviour never engaged — and FFB-Bridge no longer writes to the yoke datarefs the aircraft drives itself. Engaging command mode used to drop instantly to control wheel steering because two writers were fighting over the same value.

MOZA base ownership

• Opening FFB-Bridge no longer changes how the base feels. A managed MOZA base now keeps its own MOZA Cockpit configuration — spring, damping, friction, inertia, overall strength, and the advanced family including breakout and professional mode — until the moment you Arm. At Disarm and at quit the exact pre-flight values are written back and verified, so between flights the base is indistinguishable from one with FFB-Bridge closed. The gentle app-owned Standby spring from 1.4.1 is retired as the resting state and remains only as a fallback if the hand-back itself fails.
• A stick base is never left with nothing holding it. Some bases arrive with their own spring set to zero; handing control back to that configuration left the grip hanging under its own weight. Whenever FFB-Bridge is not armed, and at exit, a base whose own spring reads zero is held at MOZA's factory-default spring strength instead. Any non-zero spring you have configured, however soft, is kept exactly.
• While flying, a managed base always keeps at least its factory resistance. Taking over the base's configuration used to remove the damping, friction, and inertia MOZA ships with, and a spring with nothing absorbing its energy can oscillate. The factory values now act as a floor under every profile.
• New Hardware → MOZA → Base resistance controls. Resting damping, friction, and inertia minimums for the selected base, applied in every state. A profile may ask for more but never for less than you set here. On bases that carry a factory floor, going below it asks for an explicit acknowledgement first; the AY210 starts from zero.
• The AB6 cyclic stays where you leave it in helicopter mode. In a hover with force trim released, only the cyclic damper resists the stick, and the AB6's own weight pulled it forward out of your hand — bench measurement showed no friction setting could hold it. FFB-Bridge now applies a small measured counter-force that cancels the base's imbalance, plus a matching friction hold, whenever a helicopter profile is armed on this base. Flight Check's cyclic damping test now behaves as described: let go anywhere and the stick stays put. Other devices and fixed-wing profiles are unchanged.

Linux

• Arming a second time in the same session works. Force-feedback effect slots were never actually freed on disarm, so the kernel pool ran out and every later arm failed until the app was restarted. This affected every device on every distribution.
• Sticks whose Linux driver renders no periodic effects get their cues back. Where the kernel driver implements constant force and nothing else, every rumble, buffet, and mechanical cue was rejected and skipped with nothing in the interface to explain the silence. FFB-Bridge now detects that and synthesises the whole catalogue into the constant-force output automatically, without the user having to find a setting.
• A clean quit is no longer reported as a crash on the following launch.

macOS

• The app identifies itself as FFB-Bridge in the menu bar, the Force Quit list, and system dialogs, rather than as "Avalonia Application".

Devices

• Recovery when another program resets the device. A program sharing the stick can clear the firmware's effect table, after which every write from FFB-Bridge still reports success while the stick sits dead — the reported "SteamVR turns FFB-Bridge off" failure. While armed, FFB-Bridge now notices the loss and rebuilds its effects, with a bounded retry that latches with an explanatory message rather than looping.
• The Adapt-FFB-Joy adapter is recognised. The open-source adapter that puts a SideWinder Force Feedback Pro on USB (03EB:204E) is supported on Windows DirectInput, with its output update rate limited to what the adapter handles.
• Startup no longer stalls probing MOZA bases that are not attached. Scanning for supported devices used to open the firmware configuration channel before checking whether the base was present at all, costing about a second per absent base on every launch and printing an ownership error for hardware that was simply not plugged in. Absent bases are now skipped in milliseconds.

App

• The device picker scrolls. With several sticks attached, Save and Cancel could sit below the bottom of the dialog.
• Left-clicking the tray icon shows the window.
• Profile sharing works without a simulator connected. A share created with the simulator closed now carries the aircraft the profile was tuned for, taken from the profile's stored tuning context, so the website listing is complete instead of blank.

1.4.1

FFB-Bridge 1.4.1 is the stable roll-up since 1.4.0. It substantially expands hardware ownership, MOZA flight-base support, simulator correctness, helicopter force trim, force safety, and the tools used to verify a setup.

MOZA flight bases

• AB6, AB9, and AY210 are separate first-class device identities. The AB6 is recognized as 346E:1002, the AB9 as 346E:1000, and the AY210 as 346E:1001; AB6 is no longer described or treated as an alternate AB9.
• Managed firmware ownership is much broader. FFB-Bridge journals the complete supported pre-session force policy, normalizes hidden scaling, breakout, professional-mode, and directional-mechanics values where that exact model has been admitted, verifies deliberate changes, and restores the exact journaled state on app shutdown, device switch, or unrecoverable teardown. Recovery survives an interrupted prior session.
• AB9 and AY210 use their validated per-axis mechanics. Exact AB9 hardware can preserve independent pitch/roll proportions through directional firmware mechanics; AY210 keeps its validated independent HID/PID conditions. AB6 retains its validated scalar local-mechanics path until its directional firmware behavior is physically admitted.
• Force Sensing is available on exact AB9 hardware. Aircraft profiles can request a portable Force Sensing input mode independently of Manual, Hydraulic, Fly-by-wire, or Rotorcraft feel. The guided setup includes native mode ownership, four-direction force/dead-zone/travel and response-curve configuration, cogging calibration with real progress, validation before profile permission is saved, drift monitoring, and exact restore.

Autopilot Follow and force trim

• Official physical Autopilot Follow is validated on AB9 and AY210. The base moves the linked yoke or stick to the commanded pitch/roll position, isolates that motor motion from pilot input, accepts pilot pass-through, and clears stale targets automatically if the heartbeat expires. Follow speed controls how quickly the base moves; Authority controls travel.
• MOZA Follow now uses the firmware's dedicated physical-follow path at full authority. Pitch and roll targets remain independent, the inactive axis is explicitly released, and trim/force-trim continues through its separate exact-position mechanism.
• Follow uses the control position actually commanded by the autopilot. MSFS uses AP-inclusive yoke position and X-Plane uses total control ratios, with surface-deflection fallback where appropriate. This fixes centered or reversed targets while the aircraft was visibly pitching or banking.
• Legacy-device AP hold is firmer and easier to override deliberately. The target hold is now a breakout/magnetic-brake style spring at absolute strength, eliminating the weak response of the previous linear spring near center. Follow Strength sets breakout effort and Master gain still applies.
• Helicopter force trim now captures one coherent equilibrium. Releasing the force-trim control anchors the physical control and simulator datum together, avoids the old pitch-up pull, and can use the full hold strength of the device independently of the active-flight output ceiling.
• Rotorcraft controls are complete again on MOZA. The Force trim on/off assignment, one-press Trim reset, hold strength, beep trim, breakout, and firmware-local exact-position hold are available together. Auxiliary USB button boxes and other controllers can supply these assignments on Windows, Linux, and macOS.

More accurate aircraft feel and effects

• Control loading follows real aerodynamic state. MSFS and X-Plane derive feel airspeed from flight-model dynamic pressure, so control weight continues to reflect the aircraft when the pitot/static system or airspeed indicator is failed, frozen, or unpowered. Indicated airspeed remains a guarded fallback when pressure data is unavailable.
• X-Plane pressure units are corrected at runtime. qbar/Qstatic values are treated as Pascals as supplied by the live simulator, with guarded compatibility handling for installations affected by the stale documented psf units.
• Spoiler, flap, rotor, and surface telemetry was audited. Spoiler effects follow actual surface deployment, including an authoritative zero; X-Plane 11 flap fallback is restored; AP and control-surface datarefs use the correct writable/read-only roles; and unavailable telemetry degrades through explicit fallbacks.
• Helicopter vibration follows the rotor system. Rotor rumble continues while a windmilling rotor turns, two-per-rev uses rotor speed, per-blade vibration respects blade count, and retreating-blade-stall onset scales from the aircraft's sourced VNE with a dedicated tuning control.
• Engine vibration has more life at low power. Engine rumble can follow RPM and the new Idle roughness control adds a rougher low-speed texture that smooths into cruise without changing the steady control load.
• Aircraft effect levels reach the hardware without hidden attenuation. Engine rumble, bumps, buffet, and vibration follow the aircraft profile and the selected pitch/roll output limits without an additional detail cap or a duplicate hardware texture slider.

Profiles, simulator discovery, and diagnostics

• The Cessna 208B Grand Caravan has a sourced, dedicated profile. It is lighter in the pattern, deliberate in cruise, firmer in roll than pitch, and uses a 150 KIAS reference plus restrained mechanical cues derived from Caravan-specific evidence.
• Automatic profile selection is safer and explainable. An unmatched X-Plane identity keeps the current profile until a real starter or saved profile matches. Session logs and support bundles record the simulator identity, selected profile, selection reason, saved bindings, and installed profile inventory.
• MSFS detection checks the simulator itself. Reachability uses a real SimConnect OPEN handshake, ignores stale Wine/Proton listeners, reports the conflicting process and PID after repeated timeouts, searches secondary drives, and offers a manual locate path.
• Flight Check covers more of the real signal path. Guided speed and RPM sweeps exercise aerodynamic and engine-dependent effects without requiring a live flight. Leaving, passing, failing, or closing a check completes the bounded STOPALL teardown before normal tuning can return.
• Force direction has one canonical convention. The pipeline, raw backends, human-first direction verifier, resumable checkpoints, and saved legacy polarity migration agree on pitch and roll signs. Physical direction certification covers the normal raw paths for AB9, AY210, and SideWinder FFB2, plus AB9 DirectInput as a separate backend cell.

Hardware and platform compatibility

• Linux Logitech support is more reliable. Force 3D Pro, WingMan Force 3D, and G940 devices whose evdev driver exposes constant effects but no native Spring now receive synthesized centering and software-rendered periodic cues. Arm retry also reclaims scarce stale effect slots.
• Logitech G940 autocenter behavior is corrected. DirectInput autocenter writes and Spring stop/restore now use the device's expected property semantics, including the force-trim release path.
• Cross-platform auxiliary controls are supported. A separate controller can provide trim, force-trim, and other assignments across the desktop platforms while the selected force-feedback device remains the sole force output target.
• Settings are easier to navigate. General, Session, and Privacy & diagnostics are separated into clear tabs, with force safety and automatic Arm/Disarm behavior grouped under Session.

Important capability boundaries

• AB6 Follow was still gated in 1.4.1. Input/PID output, managed configuration, local effects, and scalar mechanics were supported on exact AB6 hardware. A remote guide observed no official AP motion at its tested conditioning and only weak Integrated Spring transients. That run did not establish that the official renderer was incompatible; 1.4.2's owned-hardware work later proved the same AP transport and identified E1:04 as the non-tracking AB6 path. The 1.4.1 renderer stayed disabled pending that gate.
• Force Sensing is admitted only on exact AB9 hardware. AB6, AY210, and generic MOZA-family overlays remain in Position mode unless their own capability and complete physical behavior are separately validated.
• Independent sidesticks are excluded from linked-yoke Follow. Non-MOZA devices retain their existing software-rendered AP behavior and simulator-specific authority limits.

Safer force ownership and lifecycle

• All normal flight force reaches the device through one owned path. Arm, Disarm, simulator loss, Safety Stop, device changes, and shutdown now share explicit ordering for effect cleanup, force handoff, and configuration restoration. This removes competing renderers and several abrupt force transitions.
• Disarmed controls remain supported and usable. Generic devices return to their normal native/default centering. Managed MOZA bases enter a gentle, profile-independent Standby spring after an ordinary Disarm or simulator exit, while aircraft forces, trim/follow, resistance, and vibration remain off. Simulator joystick input remains available in Standby and Disarmed states.
• Safety Stop is a true latched force cut. It gain-zeros the device, stops active effects, leaves MOZA Standby stopped, and requires an explicit re-Arm. The sustained-oscillation detector now intervenes much sooner on strong hardware; the rapid-motion guard remains independent.
• Safety policy is device-aware. Automatic motion protection remains on for strong or uncharacterized bases and is off by default for characterized legacy consumer sticks. The saved Settings control explains the tradeoff and asks for confirmation before protection is disabled.
• Bench and guided tests build force gradually. Test fixtures use their own conservative, device-class ceiling, ramp through direction changes, and restore the saved live-flight limit afterward. Strong and unassessed bases are capped at 30% during direct pitch/roll checks; normal flight can still use the explicitly selected output range.
• Fast movement near physical travel limits is softened. Edge authority is reduced only while motion is already fast and outward, then returns smoothly as the control slows or moves inward. A device-output heartbeat also detects a stalled force loop and clears output.
• X-Plane axis ownership is scoped to the active feature. Pitch and roll override flags are written only while Autopilot Follow owns pass-through and are released once when that session ends. An inactive Bridge leaves another tool's override state untouched.
• Quit separates urgent force-off from exact device restoration. The force boundary acknowledges promptly, while the complete journal restore remains visible and finishes before process exit without a false timeout or watchdog warning.

1.4.0

FFB-Bridge 1.4.0 launches the aircraft library at ffb-bridge.com and rebuilds every built-in profile from documented sources. The goal of both is the same: get the community involved in making these profiles genuinely good.

• The aircraft library is live at ffb-bridge.com/aircraft. This is the centre of this release. Every built-in aircraft now has a public page showing what the profile models, the sources behind each claim, its tuning values, and where the gaps are. You can share your own tuned profiles straight from the app (the globe button in Profiles), and a shared profile shows exactly what it changes relative to the baseline it started from. That difference is the valuable part: when someone who knows an aircraft adjusts a profile, everyone can see what they changed and why it's better. Corrections, shared profiles, and flight reports all feed the next revision of the built-in profiles.
• All 13 built-in profiles were rebuilt from documented evidence, and they need your experience. Each profile is now grounded in POH and certification data, flight-test reports, and pilot accounts: the C172 is trainer-heavy in pitch, the Extra 300 stays light from stall to VNE, the P-51D carries a strong speed load with a light per-g gradient, the transports have much firmer yokes. But documents only go so far. These profiles are built from sources, not type experience. If you have real-world time in one of these aircraft and the profile doesn't match what your hands remember, say so on that aircraft's library page or share a corrected profile. That input is exactly what the library exists for.
• X-Plane: Autopilot Follow can now be set up to 50%. While the autopilot flies, FFB-Bridge uses X-Plane's per-axis override so the simulator reads your hand input from the bridge instead of the raw joystick axis. This lets the yoke follow the autopilot with more authority without the simulator treating that motion as pilot input, and moving the yoke yourself still reaches the aircraft. This feature is new in this release and has had one day of bench testing on a single MOZA AY210: the motion is usable but not yet fully smooth, and brief counter-movements of the on-screen yoke can still appear during fast corrections. Start around 15–25%, increase to taste, and report what you find.
• MSFS: Autopilot Follow stays capped at 10%. MSFS reads the physical axis directly, so above roughly that level the simulator's autopilot and the follow cue can fight each other. Live testing showed the automatic back-off added during this cycle softens the fight but cannot stop a divergence at high authority. Saved profiles above 10% are clamped in live flight.
• MOZA AB6 with helicopters: a fix for the periodic jerk and rough centre. Two field reports traced to FFB-Bridge's background device-status traffic briefly interrupting the AB6's own damper while a rotorcraft profile was active. That traffic is now reduced to the safety-essential minimum while forces are live, and diagnostic position/torque polling runs only while the Hardware page is open. This fix could not be verified on an AB6 in-house, so reports from AB6 helicopter pilots are needed, positive or negative.
• MOZA AY210: no more unexpected reference motions, and a per-axis output envelope. The base no longer replays its startup routine during Flight Check or re-arm, Disarm is an absolute force-off on every path, and the automatic envelope is now 35% pitch and 50% roll to match the yoke's different pitch and roll mechanisms. Explicitly saved limits are preserved.
• PicoWinder is now supported out of the box on Windows and Linux, with the full SideWinder-class output range.
• A crash in the simulator now produces one bounded impact through the existing touchdown channel, within your configured force limits, and it does not repeat. A frozen or disconnected simulator behaves as before: forces off, no synthetic impact.
• The 737 stick shaker works on more add-ons. Aircraft that do not publish the standard stall-warning value get the shaker from an airborne-only angle-of-attack fallback, and the 737's shaker and buffet levels were raised to match their certified character.
• The AS 33 sailplane starter now ships with vertical-wind vibration enabled (progressive over a 1.5–15 ft/s envelope), so lift and sink are feelable. The channel stays off by default in other profiles.
• Assorted fixes: enabling Autopilot Follow can no longer soften a heavier aircraft profile's centering; engine telemetry (RPM, vibration, reverse) covers every installed engine instead of engine 1; explicit flap motion and arrival cues across the fleet.

Update from within the app, or download from ffb-bridge.com. Feedback on the rebuilt profiles, the X-Plane autopilot follow, and the AB6 fix: [email protected] or the community.

1.3.1

FFB-Bridge 1.3.1 brings the complete public 1.3.1 beta line to Stable: one place for MOZA hardware control, more accurate simulator-driven feel, clearer Flight Check diagnostics, and an absolute force-off boundary whenever the app is disarmed. Existing profiles retain their current feel; the new vertical-wind effect remains off until a user enables it.

• Disarmed now means zero force on every supported device. Disarm sets device gain to zero before STOPALL and does not replay a centering spring, damper, friction, inertia, soft stop, constant force, periodic effect, or firmware autocenter. Every backend opens at zero gain plus STOPALL and shutdown retains zero gain.
• Repeated Disarm is a physical command, not only a UI state change. Every request reissues zero gain plus STOPALL even when software already reports Disarmed. Periodic state reassertion remains silent until explicit Arm. A rejected physical STOPALL is reported instead of being presented as a successful stop.
• MOZA bases use both independent firmware output gates. Managed takeover disables force feedback and asserts torque-output inhibit before changing local damping or other coefficients. Explicit Arm enables the gates; Disarm, startup failure, and shutdown inhibit them again.
• MOZA hardware control is gathered in one place. Hardware → MOZA owns the selected base's pitch, roll, and texture limits, health telemetry, managed 13-setting session, exact shutdown restore, and isolated damping, friction, inertia, and progressive-stop tests. Aircraft-specific feel remains under Tuning. AY210 and AB9 isolated mechanics are verified on real Windows hardware.
• Force trim can use a separate Windows USB grip. A Warthog or similar controller can provide FTR and its optional toggle while the FFB base remains the actuator. Beep trim uses the same guided assignment flow. Nothing is assigned automatically.
• Aerodynamic loading and autopilot follow are more accurate. Pitch and roll loading can be balanced independently, AP-follow bank direction is corrected, Flight Check can use an unmistakable 50% isolated bench motion, and live simulator AP-follow has a hard 8% authority ceiling.
• Flap and engine feel follow more of the real aircraft. Actual flap travel drives movement, drag, buffet, and a separate arrival bump. RPM, combustion, vibration, and reverse aggregate every installed engine—MSFS engines 1–4 and X-Plane engines 0–7.
• Stall and overspeed cues can build progressively. Valid AoA, icing, stall reference, and dynamic redline data refine onset while established simulator warnings remain the fallback. Actual gear and left/right spoiler deployment, measured wheel skid, native shaker intensity, and surface condition improve their matching cues when available.
• Slew and optional telemetry fail safely. MSFS Slew is a hard force-neutral state. Each optional MSFS value is isolated; X-Plane unsupported values keep the established fallback; aircraft changes clear cached optional data. X-Plane 12 also contributes native rotor blade-slap evidence.
• Glider lift and sink can be felt without changing existing profiles. The default-off Vertical wind vibration channel uses aircraft-point environmental wind in MSFS and X-Plane 12, never aircraft climb/descent rate. It remains bounded, silent on the ground, while paused or slewing, and silent when the telemetry is unavailable.
• Windows installer builds verify startup resources. The build scripts clean RID-specific Avalonia intermediates and reject a desktop DLL missing its compiled-XAML loader output instead of packaging a build that exits before its first window appears.
• Profiles and normal armed feel are unchanged. This stable roll-up changes no existing profile automatically and includes no Pro-only simulator, licensing, model, or UI payload.

1.3.0

FFB-Bridge 1.3 is the largest release so far: first-class helicopter force feedback, managed MOZA hardware sessions, broader native device support, a self-contained Flight Check, safer startup diagnostics, and a much clearer day-to-day tuning workflow. It is a stable rollup of the complete 1.3 beta and RC series.

• Helicopters are now a first-class control system. Rotorcraft profiles combine cyclic damping, force-trim hold and release, an optional pedestal- style force-trim on/off control, continuous beep trim, AFCS follow-up trim, ETL and VRS cues, two-per-rev vibration, retreating-blade-stall buffet, skid scrape, and rotor rumble. Bell 206, H125, and generic helicopter starters cover the main trim-system families.
• Every helicopter control uses a clear guided setup. FTR and its optional on/off control use the same press-and-release modal. Beep trim asks for one physical four-way control, listens to both device formats, then saves either one POV hat or four buttons. The tested AY210 reports its hat-shaped switch as four ordinary buttons. Input preparation is shown separately with no misleading moving bar, and cancel or failed verification restores the previous assignment.
• MOZA AB6, AB9, and AY210 work as managed FFB-Bridge devices on Windows, Linux, and Apple Silicon macOS. FFB-Bridge snapshots the settings it owns, prepares and monitors a known force-feedback session, prevents stored device effects from fighting the flight model, and restores the exact prior values on a normal exit. MOZA firmware 1.1.4.6 or newer is required. MOZA Cockpit remains useful for firmware, lighting, and maintenance, but should be closed while FFB-Bridge is running.
• Native input and force paths reduce latency and remove normal-use utility dependencies. Supported SideWinder and MOZA hardware use raw HID/PID where validated, with platform-appropriate fallbacks. Held inputs such as beep trim remain continuous, and each device receives the spring, damping, friction, periodic, and one-shot rendering policy validated for its family.
• Flight Check is now the canonical hardware bench test. Direction, centering, damping, friction, trim, autopilot follow, airplane flight, and helicopter flight run from known temporary settings and restore the user's previous state afterward. Safety Stop exercises the real gain-zero, STOPALL, latch, alert, and explicit re-arm path without asking the user to provoke a runaway control.
• Hardware and aircraft feel are separated cleanly. Hardware owns device selection, the physical output ceiling, renderer/backend choice, polarity, calibration, and control assignments. Tuning starts with Profile and Master gain, then one Advanced switch and the Control system selector. Advanced reveals all relevant effect groups in one continuous page, and zero is the single off state for gain controls.
• Profiles start sensibly and follow the aircraft. A previous profile is restored when available; a fresh install starts on the Cessna 172; reported aircraft identity can then select a closer starter or a saved aircraft binding. The compact profile library adds search, sorting, import, duplicate, rename, share, and protected read-only starters.
• Startup health checks are automatic but stay out of the way. Healthy launches remain silent. Actionable device or permission failures produce a non-blocking Review-and-fix banner that opens populated Support rows. Linux verifies the active event node and complete udev coverage, including MOZA TTY policy channels; after a successful rule repair, the app recommends a restart so it can reopen the joystick with the new permissions.
• The force model is broader and more device-aware. 1.3 adds native control friction, glider airspeed feel, simulator-driven flap motion and buffet, improved Safety Stop recovery, device-normalized output ceilings, expanded roll and autopilot behavior, and a conditional 200 Hz loop for renderers that need host-side waveform or device-coupled updates.
• Hardware coverage is wider. Alongside the Microsoft SideWinder Force Feedback 2 and MOZA family, 1.3 includes the listed Logitech G940, Force 3D Pro, and WingMan Force 3D paths plus Brunner FFB-G support. Experimental unlisted joystick opt-in remains conservative and excludes wheels, gamepads, and unsuitable single-axis devices.
• Updates and shutdown are safer. Windows and packaged Linux installs can download, verify, install, and restart after explicit consent; Windows also verifies the expected Authenticode publisher. macOS opens the signed current download. Managed device restoration completes before exit, and Linux tray observers are disposed before dispatcher shutdown.

Supported simulators are Microsoft Flight Simulator 2024/2020 on Windows and Linux, and X-Plane 11/12 on Windows, Linux, and Apple Silicon macOS. Existing profiles and hardware preferences migrate forward; control assignments remain explicit and simulator keybindings are never changed.

1.2.1

• MOZA AB9/AB6 hotfix: synthesised centering spring — MOZA bases accept but don't render the DirectInput spring effect, so the bridge now reads the stick position and renders centering itself. Restores self-centering and trim's hands-off hold on MOZA hardware.

1.2.0

• Flight Check: a per-effect diagnostic page — confirm every force the bridge makes, one guided step at a time, no sim needed.
• Elevator/aileron trim sign closure (live-MSFS validated) and a one-shot polarity reset for configs that carried workaround inversions.
• MOZA AY210 yoke base joins the supported-device registry.