B&G Racing Electronics: What We're Looking For

This is a research note, not a rated review. We're documenting how B&G's racing electronics work at an engineering level and what we'd measure, before any testing on our own boat.

A racing instrument system is not a set of gauges — it is a real-time state estimator. Raw masthead and log data are noisy, biased and reference-frame-dependent; the entire value of a B&G H5000 or WTP-class system is the calibrated computation that converts them into a true-wind vector the afterguard can steer to. This note covers the sensor chain, the maths in the processor, the correction models, the network, and the criteria we'd use to evaluate a system for a Melges 40 — grounded in our general guide to race-boat electronics.

The sensor chain and what each measurement actually is

Five primary measurements feed everything downstream, and each one carries a specific error signature the processor must model out.

Apparent wind (AWA/AWS) comes from the masthead unit — either a mechanical vane-and-cups head or an ultrasonic head. This is a vector measured in the masthead's reference frame, which is the problem: at the top of a 20-metre-class rig the sensor is swinging through an arc every time the boat rolls or pitches, so its motion adds vectorially to the true airflow. The masthead also sits in disturbed air — upwash off the sail plan bends the local streamline, and rig loads twist the masthead fitting.

Speed through water (STW) comes from a through-hull paddlewheel or electromagnetic log. Critically it must be through water, not over ground: true wind is defined relative to the water, so feeding speed-over-ground (SOG) from GPS contaminates the true-wind solution with tidal current. The log reads in a thin, hull-disturbed boundary layer and is sensitive to heel, to fore-and-aft trim, and — if it is not exactly on centreline — to which tack you are on.

Heading comes from a fluxgate or solid-state (MEMS/fibre) compass. It must be heading (where the bow points), distinct from course-over-ground (COG, where the boat tracks). The angular difference between them, once current is removed, is leeway.

Attitude and rates — heel, trim, and the roll/pitch/yaw rates — come from a rate gyro/accelerometer package. On B&G this is the H5000 3D Motion Sensor or the inertial platform inside a WTP3.

Depth from the transducer is tactical (avoiding the beach on a layline) rather than part of the wind solution, but it shares the same bus and time base.

The true-wind computation

At its core the true-wind solution is straightforward vector subtraction. Resolve apparent wind into fore-aft and athwartships components, subtract the boat's velocity-through-water vector, and recombine:

• TWS = √( (AWS·cos AWA − BSP)² + (AWS·sin AWA)² )
• TWA = atan2( AWS·sin AWA , AWS·cos AWA − BSP )

atan2 (rather than plain atan) is what resolves the angle correctly through all four quadrants — the difference between a sane true-wind angle and a nonsensical one running deep downwind. True wind direction (TWD) then adds heading: TWD = Heading + TWA (with sign for tack). Because TWD is the difference of two large quantities near the beat and the run, small errors in AWA, BSP or heading are amplified — a one-degree masthead misalignment can move TWD by several degrees, which is the difference between calling a header and calling a lift. This sensitivity is precisely why the correction layers below matter more than raw sensor accuracy.

A subtlety that separates a properly set-up system from a naive one: leeway cannot simply be added to apparent wind angle. A 5° leeway angle does not shift TWD by 5°; because it enters the vector geometry rather than the angle directly, the net effect on TWD is closer to 1°. B&G resolves the boat's velocity vector using both STW and modelled leeway, which is why the leeway model has to be right.

Motion correction — the reverse pendulum

This is the single biggest differentiator between a cruising instrument and a Grand Prix one. In any seaway the masthead accelerates through metres of travel; those accelerations project onto the wind vane and corrupt AWS and AWA at exactly the frequencies of pitch and roll. Uncorrected, TWD oscillates and the numbers are unusable in waves — which is when tactical data matters most.

B&G's approach corrects in two stages. First it applies a static heel-and-trim correction: an inclined masthead unit measures the horizontal wind component skewed by the tilt angle, a pure trigonometric correction from the heel and trim sensor. Then it applies dynamic rate correction: roll, pitch and yaw rates from the motion sensor are combined with the known geometry of the rig to compute the masthead's instantaneous velocity, which is then subtracted from the apparent-wind vector. Because the rate sensor sits at the mast base rather than the masthead, and the mast length is known, the processor runs what is effectively a reverse-pendulum calculation — the masthead velocity is the base rotation rate multiplied by the height. WTP-class systems refine this further by fusing a masthead accelerometer with the base gyro and applying bus-latency compensation so that motion and wind samples are aligned in time. Get the time alignment wrong and the correction adds noise instead of removing it.

For a Melges 40 this is not optional. It is a light (nominally ~3,250 kg, verify against class documentation), powerful boat with a tall square-top rig that pitches and rolls hard in a chop; the masthead motion signal is large relative to the true wind. Any evaluation has to distinguish genuine inertial motion correction from a heel-only trig correction that leaves the dynamic accelerations in the signal.

The correction models that make TWD trustworthy

Beyond geometry, the processor carries empirical tables that model the physics the equations don't capture.

Upwash and mast twist. The sail plan deflects the airflow at the masthead so that, upwind, AWA reads further aft than the free-stream angle — the wind appears more astern than it is. Working against this, rig load (mainsheet and running backstay tension) twists the mast and swings the sensor arm to windward, biasing the reading forward. The two partially cancel but not cleanly, and the residual varies with wind speed and point of sail. B&G handles this with upwash tables carrying separate correction values for beating, reaching and running across a grid of true-wind speeds (typically 5/10/15/20/25/30 knots). Downwind there is a second effect: the true-wind speed over-reads because air accelerates over the masthead in the accelerated flow field, and corrections of roughly +12% to −15% are applied depending on angle and speed. Vertically-oriented wind sensors reduce both upwash and downwind acceleration error by presenting the vane to cleaner flow, which is why performance boats favour them.

Leeway. The keel generates side-force by operating at a small angle of attack — the leeway angle — and from wing theory that lift, and hence the leeway needed to balance the sail side-force, scales with heel and inversely with the square of boatspeed. The standard model B&G uses is:

• Leeway = K · Heel / BSP²

with heel and leeway in degrees, BSP in knots, and K a boat-specific coefficient typically between 9 and 13 (it can vary with wind speed). K is not a number you can look up — it is derived on the water, classically by sailing upwind on a tack while logging heading, heel, STW, COG and SOG, then bearing away and running dead downwind (where leeway is zero) to isolate the current, so the residual heading-minus-track upwind is true leeway. For our boat this coefficient will need to be measured, not assumed.

Speed-through-water calibration. A single scale factor is never enough on a race boat. The log reads differently port versus starboard if it is off centreline, because the flow angle across the paddlewheel changes with tack; and the reading drifts with heel as the hull's boundary layer changes over the sensor. A proper set-up uses a multi-point speed table (not one linear coefficient), a tack-to-tack offset, and a speed/heel correction, calibrated against a reliable ground reference on a no-current day. Marine growth on the paddlewheel is a slow, insidious source of under-reading that quietly poisons the whole true-wind solution, so the transducer has to spin freely and stay clean.

Damping. All of this is filtered before display. B&G's newer digital masthead units transmit at 10 Hz but apply several seconds of internal damping; the processor adds further, ideally adaptive, smoothing so that steady-state numbers are stable while the system still responds through a tack. Too little damping and TWD is jumpy and untactical; too much and it lags the shift you're trying to read. Where that time constant is set — and whether it adapts to boatspeed and manoeuvre state — is a real evaluation point.

Network architecture, latency and time base

The physical backbone matters because every correction depends on samples being aligned in time. Legacy B&G ran the proprietary FastNet serial bus (28800 baud); current H5000 and WTP systems use NMEA 2000 (CAN) for sensor traffic plus Ethernet for high-bandwidth display and configuration data, with FastNet interfaces available to bridge older sensors onto the CAN network. Tactical outputs — SailSteer, laylines, start-line and target data — are computed and pushed at up to 10 Hz.

The engineering risks are latency and datum consistency. If wind, heading and rate samples arrive with different bus delays, the motion correction subtracts a masthead velocity that no longer matches the wind sample it is correcting — hence B&G's explicit bus-latency compensation. Time-stamping and a coherent update rate across CAN and Ethernet are therefore a genuine evaluation criterion, not a spec-sheet footnote, and they matter more the harder the boat is being pushed.

Integration with navigation and performance software

The processor is the sensor authority; the navigator's software — for us, Expedition — is where polars, routing and current modelling live. The clean division is: instruments own the calibrated true-wind and boatspeed datum, the nav computer owns the strategic layer. But the seam is a classic failure point: Expedition can carry its own calibration and its own current model, and if it applies a second layer of wind correction on top of the processor's, or resolves current differently, the two disagree and the crew loses trust in both. Pre-start, current derived live from instruments is unreliable (there is no steady reference), so navigators typically set current manually via "what-if" functions rather than trusting the computed set and drift until the boat settles. A key question for any system is how cleanly it exports a single, already-corrected datum so corrections are applied once, in one place.

Power, environment and serviceability

At Grand Prix level none of this runs on a whim: the processor, colour displays, autopilot and masthead electronics draw continuously, and the masthead unit and its cable run are the least serviceable items on the boat — a masthead fault means going up the rig. So power budget at the DC bus, connector sealing and strain relief at the mast base, transducer access in the bilge, and firmware/configuration recoverability all sit on the evaluation list next to raw data quality. A system that produces beautiful numbers but strands you with an unreachable masthead fault mid-regatta has failed at the thing that matters.

What we'd look for on our boat

Our evaluation criteria, in priority order:

• Wind-data integrity — masthead update rate, internal versus processor damping, and above all whether motion correction is genuine inertial compensation. The test is TWD stability through a tack and in a seaway.
• Calibration depth — multi-point speed table, tack-to-tack log offset, speed/heel correction, upwash tables by wind speed and point of sail, and a measured leeway K.
• Network latency and time base — coherent sample timing across CAN and Ethernet so corrections stay aligned.
• Integration — a single corrected datum exported to Expedition with no double-correction.
• Display legibility and refresh under load, autopilot integration, and crew usability.
• Power draw and serviceability — DC budget, and masthead/transducer access and recoverability.

The aim is a true-wind datum the crew and navigator trust and steer to, not the system with the most features.

Who it's for

Sailing-specific instrument systems earn their cost on boats where TWD drives every tactical call and where the boat's own motion is large enough to corrupt the wind signal — exactly the Grand Prix keelboat case. Club racers can run far simpler kit; a canting-keel, twin-rudder one-design like the Melges 40, pushed hard in waves, sits at the demanding end where inertial motion correction and a fully calibrated pipeline stop being luxuries.

Next step

When we have genuine data from a system on our own boat — TWD stability logs through tacks, a measured leeway coefficient, calibrated speed tables — we'll publish a field review with honest findings: what the numbers actually did, what annoyed us, and whether we'd choose it again. Until then, this stays a research note. See also our notes on Sailmon displays and Expedition software.