The pattern and sensing optics form a pair
In transmitted-light reading, source and detector are on opposite sides of the pattern; in reflected-light reading they share a side. Some readers use a stationary scanning grating, while others use direct imaging, detector arrays or diffraction. A separate scanning reticle is not required in every optical encoder. Match track pitch, reading radius, wavelength, sensing aperture, film side and working gap to the actual architecture.
Clear fraction and optical contrast
At a stated reading radius, clear fraction κ = clear width/pitch. A 50% clear fraction gives equal clear and opaque widths; for an ideal binary grating it maximises the fundamental Fourier component. It does not by itself guarantee 50% electrical duty cycle, which also depends on illumination, diffraction, detector aperture and thresholds. Measure optical contrast using the actual collected levels: C = (Imax − Imin)/(Imax + Imin).
Specified wavelength matters for both clear-region transmission and coating leakage or reflectance. Visible inspection does not replace the near-infrared or other wavelength measurement used by the reader.
Phase-shifted sensing and differential channels
In an ideal four-field reader, sensing phases of 0°, 90°, 180° and 270° produce opposing signals. Subtracting opposite fields can reduce a common background term and even harmonics when gains, illumination and phase spacing are matched. Real mismatch leaves residual offsets and distortion. The quarters are phase offsets within an optical period, not four readheads spaced around the shaft.
Recover phase with corrected sine and cosine
For one signal period, use corrected channels S = A sin φ and C = A cos φ. Phase is atan2(S,C); atan2 retains the quadrant that a simple S/C ratio loses. Unwrap phase and combine it with the period count or absolute address to recover travel. For physical signal period p, fine displacement is pφ/(2π), subject to the chosen phase origin and direction.
Signal defects create within-period errors
Unequal amplitudes, DC offsets, quadrature error and harmonic distortion alter the sine/cosine locus and produce cyclic phase error. Gain imbalance commonly creates a component twice per signal period; offsets can create a once-per-period component. These frequencies are relative to the signal period, not automatically once or twice per shaft revolution. Record corrected and uncorrected waveforms with position.
| Observed feature | What to investigate |
|---|---|
| Unequal amplitudes | Optical balance, detector gain and gain correction. |
| Shifted signal centre | Background light, channel offsets and subtraction. |
| Non-quadrature phase | Sensing-field geometry, alignment and phase correction. |
| Distorted or noisy waveforms | Pattern contrast, aperture, contamination, bandwidth and noise. |
Resolution does not scale away phase noise
For small phase noise δφ, displacement noise is approximately pδφ/(2π). Increasing the displayed subdivision K reduces the count increment p/K, but does not reduce a fixed physical phase error. The same error occupies more output counts as K increases. Filtering may lower noise while adding delay, so validate bandwidth and dynamic tracking as well as stationary readings. There is no universal gain-imbalance limit determined by interpolation factor alone.
Geometric, moiré and diffractive readings
Geometric readers modulate collected intensity with the moving pattern. Nearly matched gratings can create a moiré envelope; for parallel pitches p1 and p2 its ideal beat period is 1/|1/p1 − 1/p2|. Diffractive readers use selected diffraction orders and interference, so the detected signal period may differ from the manufactured pitch. These mechanisms depend on wavelength, gap and alignment; a small reported increment alone does not identify the readhead technology.