Calibrators¶

Polynomial calibration¶

Polynomial describes the transfer curve as polynomial coefficients, ordered from x⁰ upward — Polynomial([c0, c1, c2]) means eng = c0 + c1·raw + c2·raw²:

voltage = Y.FloatParameter(
    system=eps,
    name="bus_voltage",
    units="V",
    encoding=Y.uint12_t,
    calibrator=Y.Polynomial([0.0, 0.00805]),   # eng = 0.00805 * raw
)

A linear scale-and-offset is simply a two-coefficient polynomial.

Piecewise interpolation¶

Interpolate describes the curve as a set of points; values in between are interpolated linearly. xp holds the raw coordinates (increasing), fp the corresponding engineering values:

temperature = Y.FloatParameter(
    system=eps,
    name="battery_temp",
    units="degC",
    encoding=Y.uint8_t,
    calibrator=Y.Interpolate(
        xp=[0, 64, 128, 192, 255],
        fp=[-50.0, -10.0, 20.0, 45.0, 60.0],
    ),
)

This suits thermistors and other sensors with non-linear response, where the calibration comes from a lookup table.

Computing calibrators¶

Since the model is a Python program, calibration data does not have to be transcribed by hand. At its simplest, writing coefficients as arithmetic keeps the derivation visible in the source:

# 10-bit ADC, 3.3 V reference, 20 mV/unit sensor
calibrator=Y.Polynomial([0, 3.3 / 1023 * 0.02])

And when the transfer function of the sensor circuit is known, the calibrator can be derived at build time — sample the function over the raw range and either feed the points straight into Interpolate, or fit a polynomial with a library such as NumPy:

import numpy

def pt1000_temperature(raw):
    """Transfer function of the PT1000 readout circuit."""
    v_out = raw * 3 / 1023
    r = (2310 + 462 * v_out) / (3 - (0.231 + 0.0462 * v_out))
    return (r / 1000 - 1) / 3.85e-3

xp = list(range(0, 1023, 10))
fp = [pt1000_temperature(x) for x in xp]

coef = numpy.polyfit(xp, fp, 2).tolist()
coef.reverse()   # polyfit returns highest degree first; Polynomial wants x^0 first
calibrator = Y.Polynomial(coef)

Note the reverse(): NumPy’s polyfit orders coefficients from the highest power down, while Polynomial expects them from x⁰ upward. Libraries used this way are build-time tools only — PyMDB itself has no such dependencies, and nothing of them remains in the exported XTCE.