Dual Pedal Channels: Building Plausibility Windows from Supplied Curves

Write the nominal channel relationship before choosing limits

Tabulate paired channel A and channel B outputs at the same independently measured pedal positions. Include electrical endpoints, mechanical stops, intentional plateaus and every curve breakpoint. If one channel rises while the other falls, preserve that direction in the mapping rather than reversing data during presentation. The nominal relationship may be linear over one interval and piecewise elsewhere. A diagnostic residual should be zero, or another declared baseline, for the intended paired curves. It should not be defined from a convenient production sample whose gain and offset already include unknown assembly error.

Name voltage references for both channels. Signals that share excitation or ADC reference can move together and remain mutually plausible while both misrepresent pedal position. The independent mechanical coordinate is therefore essential during development. Once the mapping is released, the controller may evaluate only electrical channels in service, but the validation evidence must show what common physical input the pair was intended to represent.

Remove intended gain and offset from the diagnostic residual

For a linear interval, predict channel B from channel A using the supplied gain and offset, then subtract that prediction from measured B. For a nonlinear relationship, use the approved table or function and retain its valid domain. Do not refit gain and offset separately for every tested unit before checking plausibility; that process can remove the manufacturing or assembly variation the diagnostic is meant to expose. Coefficients should have a clear origin, revision and rounding rule. Sampling must be time-aligned when the pedal is moving because a small delay appears as a channel disagreement proportional to travel speed.

Calculate one mapped point and retain its context

At an illustrative travel point, suppose channel A is 0.800 V and the released relationship predicts B = 1.5A + 0.25 V. The predicted B output is 1.450 V. If the simultaneous measured B value is 1.460 V, residual is +10 mV. A ±20 mV window would pass this point, but that example does not establish a suitable production or diagnostic threshold. The allowed band must also cover justified tolerances and uncertainty without becoming so wide that the required fault distinctions disappear.

Repeat the arithmetic at neighbouring points rather than extrapolating this result. If local slopes differ, the same position mismatch produces a different voltage residual. Near saturation, the mapping may become insensitive to movement and should be excluded or governed by a separate endpoint rule. Quantization can make a narrow band alternate between adjacent codes; use the real ADC and filtering model when converting an analog residual into controller counts.

Build upper and lower envelopes from identifiable contributors

At each travel station, propagate channel A and B output tolerances, mapping coefficient uncertainty, mechanical phase difference, ADC uncertainty and permitted timing skew into the residual coordinate. Preserve sign where a contributor pushes only one side of the band. Shared excitation movement may cancel partly in the residual; do not count it as two independent full errors. Conversely, shared return impedance can affect the channels unequally if their currents or references differ. Use worst-direction sums for bounded unknowns and statistical combination only where distributions and correlations are supported.

The completed envelope can vary with position, temperature and supply state. Store the bounds as a reviewed table or function, including interpolation behavior. A percentage-of-full-scale window is easy to communicate but may be technically weak in a region of low local slope. Express the physical position consequence alongside voltage residual during design review so engineers understand what the diagnostic can and cannot distinguish.

Separate static plausibility from moving-pedal timing

Measure stationary residual after a declared settling time, then evaluate controlled ramps at several speeds. A residual that is small when held but increases with speed suggests sampling or filter phase rather than static track mismatch. Compare increasing and decreasing motion at the same mechanical positions. Direction-dependent separation can arise from shared carrier clearance, wiper friction or track-to-track registration. Do not average the two branches before checking the integrator's requirement. A controller may need an independent dynamic allowance, but that allowance should not enlarge the stationary band and hide a real offset.

Retain raw channels before software normalization. A plausibility algorithm can appear stable because both channels use the same erroneous position estimate or are filtered by the same delayed state. Development plots should include independent position, velocity, excitation and both unscaled voltages on one timebase. This evidence exposes whether the residual belongs to the card, mechanism, wiring or data acquisition.

State what channel agreement cannot detect

Agreement between the mapped signals does not prove correct pedal position. A shifted shaft, common carrier movement, shared supply loss or common ADC reference error can preserve their relationship. List these dependencies explicitly and assign other system measures where needed. The resistor card can provide separated printed paths and terminals according to a drawing; it cannot by itself guarantee diagnostic independence. Claims about safety integrity, coverage or vehicle response require a broader analysis and evidence outside this calculation.

A useful fault exercise changes one path at a time and then challenges selected common nodes. Examples include bounded input loading, a controlled series resistance at one terminal, supply variation within the approved range and a mechanically referenced phase shift. Such work must follow the responsible laboratory's safe procedure. The purpose is to observe residual direction and detection timing, not to simulate every possible field failure.

Validate the released window without fitting the test data to pass

Freeze coefficients and bounds before evaluating the validation set. Exercise multiple specimens through the complete usable travel, both directions, relevant stationary temperatures and specified supply states. Log every excursion with its raw channel values, position and timing. Repeating a failed point after changing a coefficient is development, not confirmation of the original window; preserve both revisions. Include independent reference checks capable of revealing common fixture movement or source drift. Review regions adjacent to breakpoints closely because interpolation and coefficient switching can create discontinuities that endpoint testing misses.

Separate card acceptance from controller diagnostic behavior. Incoming electrical measurements may verify channel resistance and transfer under a test load, while assembled-system work verifies mapping, timing and reaction. The documentation should name which owner approves each result. This prevents a component supplier from appearing to certify the complete accelerator system and prevents the integrator from expecting a bare-card test to reproduce installed filtering.

Keep curves, mapping and diagnostic policy under revision control

The interface package should contain both terminal schematics, paired nominal curves, valid travel, mechanical datums, mapping coefficients, window tables, ADC references and sampling assumptions. Identify saturation, open-circuit and short-circuit handling separately from ordinary plausibility. When artwork, wiper geometry, carrier, terminal assignment, filter or coefficient changes, reopen the relationship review. A curve image without numerical points and a named interpolation rule is insufficient for reproducible calculation.