chromatic_confocal_height — INTERFEROMETRY chromatic op

데이터 종류: sweepmeasurement

호출: import interferometry; interferometry.chromatic_confocal_height(spectrum, wavelength_start_nm=500.0, wavelength_step_nm=0.5, dispersion_um_per_nm=0.2, reference_wavelength_nm=600.0, mode='gaussian', subtract_background=True, min_visibility=0.3, min_peak_bins=2.0, max_carrier_fraction=0.1)(또는 opsinterferometry.get("chromatic_confocal_height"))

사용법

공초점 반사 스펙트럼 하나로부터 표면 높이를 —— 파장이 높이입니다.

> 아래 상세 설명은 원문입니다 —— 요약과 제목은 번역되어 있습니다.

The inverse of :func:chromatic_confocal_simulate. Find the peak wavelength

of the spectrum, then::

height = (lambda_peak - reference_wavelength_nm) * dispersion_um_per_nm

No scan, no moving parts, one spectrum per point — which is why this family

reaches sampling rates a z scan cannot, and why it is limited to the axial

range the objective's chromatic spread covers.

The four *mode* estimators are :data:ESTIMATORS, identical to

:func:csi_peak_position's and sharing its implementation. `"gaussian"` is

exact for a Gaussian confocal response (measured 0.0e+00 to 3.6e-15 um),

*including when the peak is narrower than one bin and when it sits two bins

from the band edge* — the logarithm of a sampled Gaussian is a parabola

whatever its width, and the fit is local. That exactness is a property of

noiseless data only, which is what *min_peak_bins* is about.

spectrum: 1-D non-negative intensities across the spectrometer.

wavelength_start_nm / wavelength_step_nm: the spectrometer axis.

dispersion_um_per_nm: the axial chromatic calibration constant.

reference_wavelength_nm: the wavelength that focuses at height 0.

subtract_background: subtract the spectrum's median before locating the peak.

On by default because a pedestal drags the `centroid`

estimator toward the middle of the band exactly as it

drags the CSI centroid toward the middle of the scan.

min_visibility: refuse a spectrum whose peak prominence

`(max-median)/max` is below this — a flat spectrum has

no focused wavelength and its argmax is noise.

min_peak_bins: refuse a peak whose full width at half maximum spans

fewer than this many bins. Undersampling does not break

the noiseless algebra, but it destroys its noise

rejection. Measured at 1 % noise (1000 peak counts,

sigma_n = 10, 100 trials), RMS error in locating the

peak against the number of bins across its FWHM:

0.5 bins -> 0.256 nm

1.0 bins -> 0.137 nm

2.0 bins -> 0.010 nm

4.0 bins -> 0.030 nm

8.0 bins -> 0.118 nm

Two bins is the optimum and a half-bin peak is 25x

worse, from a spectrum that looks perfectly healthy —

hence the default of 2. Note the curve turns around

again: a very *broad* peak is also bad, because the

three-point fit then sits where the curvature is tiny

and noise dominates it. "More samples is better" is

false here and this operator does not claim it. Set to

0 to disable the check if you know your data is

clean.

max_carrier_fraction: refuse a spectrum whose dominant alternating component

sits above this fraction of the Nyquist frequency. A

confocal response is one smooth peak and all of its AC

content is at low frequency (measured: dominant bin at

0.010 of Nyquist for a 4 nm peak, 0.010 for a 1 nm peak,

0.015 with 5 % noise). A z-scan interferogram put in

here instead sits at 0.333 — its fringe carrier — and

without this check its carrier's argmax would come back

as a focused wavelength and therefore as a plausible,

finite, wrong height. This is the guard that lets the

two 1-D families share one type pool safely. Pass 0 to

disable.

Returns the height as a float, in micrometres. It may be negative — a height

is signed, unlike a time-of-flight distance.

Raises `ValueError`: a non-1-D / empty / too-short (< 3) / non-finite /

complex / masked *spectrum*, a *spectrum* over :data:MAX_SCAN_POINTS

elements (checked before the float64 promotion), any negative intensity

(a spectrometer cannot read negative light; clip explicitly if this is a

pre-subtracted spectrum), a non-positive step / dispersion / reference

wavelength, an unknown *mode*, a peak prominence below *min_visibility*, a

peak narrower than *min_peak_bins*, and a peak on the first or last bin

(the surface is outside the calibrated axial range, which is

`+-(band_nm/2) * dispersion_um_per_nm` about the reference).

자세한 사용 가이드

coherence_scanning 패밀리 가이드

참고(샘플 데이터·문헌)

• 샘플 데이터 카탈로그(DL URL / 라이선스) —— 2-D 는 skimage.data(BSD/public)+ 합성, 3-D 는 실데이터 소스(Stanford/PDS 등)의 DL URL.

• 연산자의 내력·참고문헌 —— 이 연산자 족의 바탕이 된 연구/기법의 출처.

• 알고리즘의 정전(저자·연도)과 용도는 위의 패밀리 사용 가이드에 적혀 있습니다.

실행 가능한 예제(이 연산자를 실제로 호출하는 검증된 샘플)

coherence_scanningpy -3.11 examples/coherence_scanning.py

타입이 이어지는 다음 연산자(measurement 를 입력으로 받는 것)

같은 카테고리(chromatic)


*Provenance: interferometry.py — INTERFEROMETRY 연산자 레지스트리. 이 op 노트는 tools/opdocs.py md 가 자동 생성합니다(직접 편집하지 마세요).*

© 2026 Kazufumi Furuse — Fullseye operator documentation. Licensed under Apache-2.0.