Rasnik: 3-Point Optical Alignment System

• Rasnik is a 3-point optical alignment system defined by a back-illuminated coded mask, a lens/objective, and a pixel sensor that together enable absolute, drift-free displacement measurements.
• It employs 2D FFT-based image analysis to extract key parameters (x, y, scale, and rotation) with sub-nanometer to picometer-class resolution, ensuring robust performance against noise and systematic errors.
• The system can be extended into N-point networks for large-scale applications and adapted to challenging environments (e.g., cryogenic, vacuum) by leveraging modified optics and calibration techniques.

Rasnik is a 3-point optical alignment and displacement monitoring system in which the measured geometry is defined by three physically separate optical elements: a coded mask, an imaging element, and an image sensor. The name denotes the Red Alignment System Nikhef, and the concept originated in CERN detector alignment before evolving into CMOS-based, software-defined metrology with sub-nanometer and picometer-class sensitivity. In its classical form, Rasnik images a back-illuminated coded mask through a positive lens onto a pixel sensor; in later variants it has been extended to microscope-objective implementations, real-time FFT-based reconstruction pipelines, and cryogenic multi-point alignment networks for accelerator systems (Graaf et al., 2021, Graaf et al., 2023, Koushik et al., 3 Oct 2025).

1. Historical emergence and system architecture

Rasnik was first developed in 1983 for muon chamber alignment in the L3 experiment at CERN, and the CCD-Rasnik system was developed in 1993 for monitoring the alignment of the ATLAS Muon Spectrometer. Large-scale deployment followed: about 8000 Rasnik systems were operated in ATLAS. This historical trajectory is central to the system’s identity, because Rasnik was conceived not as a laboratory-only displacement probe but as a distributed alignment monitor for large detector infrastructures (Graaf et al., 2021).

The canonical Rasnik geometry consists of three elements placed along an optical line: a back-illuminated coded mask, a lens or objective, and a pixel sensor / camera. The image formed on the sensor encodes the relative geometry of those three elements. In this sense, Rasnik is a genuine 3-point alignment system: the measurement is not reducible to a single target coordinate, because the inferred displacement depends on the mutual position and orientation of all three components.

Several later implementations preserve this 3-point logic while changing the optics. In the C$^3$ accelerator proposal, for example, the classical mask-plus-lens arrangement is replaced by a monochromatic point-like light source, a zone plate, and a CMOS image sensor, but the measurement remains a 3-point relation between source, optical element, and sensor (Graaf et al., 2023).

2. Optical measurement principle and measured degrees of freedom

In the classical system, the optical axis runs from the center of the sensor through the optical center of the lens to the mask. Rasnik image analysis determines the coordinates on the mask at which that optical axis intersects the coded pattern. The standard outputs are four parameters: $x$, $y$, $S$, and $\theta_z$. Here $x$ and $y$ are the mask coordinates of the optical-axis crossing point, $S$ is the image scale, and $\theta_z$ is the differential rotation of mask and sensor around the optical axis. The image scale is defined as

$$S = \frac{\text{image distance}}{\text{object distance}} = \frac{b}{a}.$$

If the lens and sensor are fixed together, Rasnik becomes a monitor of the coded mask’s 4D position: transverse position through $x$0 and $x$1, longitudinal change through $x$2, and in-plane rotation through $x$3. The system is explicitly described as having no cross coupling between $x$4, $x$5, $x$6, and $x$7, while the perpendicularity of $x$8 and $x$9 is defined by the mask itself (Graaf et al., 2021).

The 2025 formulation states the same measurement scope in operational terms: Rasnik is designed to measure transverse translation in $y$0, transverse translation in $y$1, longitudinal translation in $y$2, and roll rotation about the optical axis, with the caveat that the roll of the lens itself is not determined. This later description also emphasizes a defining asymmetry relative to many interferometric readouts: Rasnik is especially sensitive to motions perpendicular to the optical axis (Koushik et al., 3 Oct 2025).

3. Coded masks, absolute coordinates, and reconstruction algorithms

A central Rasnik innovation is the use of a coded chessboard or ChessField mask. The fine checker pattern provides dense black-white contours for precision interpolation, while a coarser coding scheme removes periodic ambiguity and enables absolute position measurement. In the 2025 implementation, every block is a $y$3 structure in which the ninth row and ninth column carry 8-bit binary values; there are $y$4 ordinary chessboard squares between coded values, and for the masks used there $y$5 (Koushik et al., 3 Oct 2025).

This architecture gives Rasnik a distinctive metrological property: no cumulative drift. The image itself contains both a periodic fine pattern and an absolute code, so position can be reconstructed after shutdown, restart, dropped frames, or long travel. The dynamic range is therefore set by mask size rather than by one chess period (Graaf et al., 2021).

Two major reconstruction frameworks are described in the literature. The earlier system uses the image-analysis program SOAP, which transforms the raw image into a 2D Fourier image, identifies the primary Fourier peaks to determine pattern frequency and possible rotation, extracts image shift from the complex phases, uses secondary or harmonic peaks for refinement, and then decodes the coarse code lines (Graaf et al., 2021). The later framework, RasCal, presents a fuller acquisition-and-control stack around the same basic principle. It begins with image acquisition, ROI selection, binning if needed, and Hann windowing, then applies a 2D FFT,

$y$6

and analyzes the peak structure generated by the periodic mask. The paper writes the odd-harmonic peak locations as

$y$7

with phase-derived fine offsets

$y$8

Absolute coordinates are then reconstructed by combining decoded integer indices, fine offsets, measured rotation, and pixel-to-physical scaling:

$y$9

with

$S$0

RasCal also defines a geometric consistency statistic,

$S$1

for the primary FFT peaks; $S$2 corresponds to perfectly perpendicular equal-magnitude peaks, and $S$3 is used as a distortion indicator (Koushik et al., 3 Oct 2025).

4. Precision limits, shot noise, and linearity

The 2021 performance study identifies the principal precision limit in unambiguous terms: once environmental and mechanical disturbances are sufficiently suppressed and the optics are diffraction limited, Rasnik spatial resolution is limited by quantum fluctuations in the applied light. The measured pixel-content fluctuation law is

$S$4

where the constant term represents electronic noise and the signal-dependent term represents photon statistics. The study supports this interpretation by showing the expected $S$5 degradation when the light level is halved and the expected frame-rate scaling of white image noise (Graaf et al., 2021).

Representative implementations illustrate the progression from sub-nanometer operation to picometer-class ASD floors:

Implementation	Context	Reported performance
SolidRas	Vacuum, suspended rigid assembly	$S$6 in $S$7
MicroRas	$S$8 objective at $S$9	$\theta_z$0
RasCal / VATIGrav	Locked proof mass at $\theta_z$1	approaching $\theta_z$2

The SolidRas system reported a best floor of $\theta_z$3 in both $\theta_z$4 and $\theta_z$5, while MicroRas, using a Newport M-20X $\theta_z$6 microscope objective, reached $\theta_z$7 at $\theta_z$8. The later RasCal framework reports $\theta_z$9 displacement sensitivity in the VATIGrav setup, with ASD below $x$0 between $x$1 and $x$2 and approaching $x$3 at high frequencies (Graaf et al., 2021, Koushik et al., 3 Oct 2025).

Statistical sensitivity, however, does not exhaust Rasnik performance. The 2025 study highlights a major systematic limitation: periodic linearity error from discrete Fourier phase extraction. The reported amplitude can reach $x$4; at $x$5 the error is $x$6 to $x$7 peak-to-peak, at $x$8 it drops to about $x$9, and at $y$0 the periodic structure is minimal. The paper models the dynamic manifestation of this effect as

$y$1

making explicit that a static periodic nonlinearity appears as velocity-dependent noise during motion (Koushik et al., 3 Oct 2025).

5. N-point chaining, cryogenic variants, and accelerator alignment

Although Rasnik is fundamentally a 3-point system, it can be extended into an N-point alignment system by daisy-chaining or “leap frogging” many overlapping 3-point units. This idea is developed for the C$y$2 accelerator, where each intermediate module carries all three functional elements needed for participation in adjacent triplets. In that proposal, each Stick contains a fixed CMOS image sensor chip, a milled pattern forming a zone lens, a mounted laser diode, and a mechanical interface to the accelerator structure or quadrupole (Graaf et al., 2023).

The C$y$3 environment motivates a specific Rasnik variant because alignment must function in ambient air, vacuum, and boiling liquid nitrogen at about 79 K, while the conclusion also mentions operation at 77 K. The key optical difficulty is the change in refractive index, with $y$4. A conventional lens that is in focus in liquid nitrogen will not remain in focus in vacuum, and for spans exceeding 20 m the required lens size becomes impractical. The proposed solution is the RasDif architecture, in which the back-illuminated coded mask is replaced by a monochromatic point-like light source, the lens is replaced by a zone plate, and the sensor records the resulting diffraction pattern. The measured quantity is the sagitta, defined as the transverse offset of the middle optical element relative to the line connecting the two end elements (Graaf et al., 2023).

Chain metrology introduces a characteristic uncertainty growth law. For a chain of $y$5 Rasnik systems, the paper states that the worst-case reconstructed spatial error scales as

$y$6

In a Monte Carlo study with $y$7, the largest error occurred near the middle of the chain when end plates were fixed. For a SuperSector containing 728 accelerator structures, two continuous Rasnik chains yield an estimated largest uncertainty in the middle of each chain of approximately

$y$8

and additional Rasnik systems at sticks 1, 364, and 728 would reduce this by another factor of $y$9 (Graaf et al., 2023).

The same work outlines a calibration procedure using a Calibration Station and three special Sticks. After rotating the station configuration by $S$0 around the $S$1 axis, the image shift on the right-hand sensor equals $S$2, allowing calibration of the source, zone plate, and sensor coordinates relative to the Stick’s mechanical reference. For the Quarter Cryogenic Module, equipped with 4-fold Rasnik chains placed at both sides and a total of 8 Sticks, the paper estimates that the relative positions of the two accelerator structures and the quad will be known within roughly $S$3 after mounting the calibrated Sticks (Graaf et al., 2023).

Cryogenic feasibility is supported by component tests. Laser diodes were operated immersed in LN$S$4 at 10 mA, typically requiring the supply voltage to be raised from around 2 V at 293 K to 6 V in LN$S$5. The paper identifies three “LN$S$6-proof” laser diode types: Laser Components ADL65074TR, Laser Components ADL65055TL, and ROHM RDL65MZT7. It also reports that a Microsoft HD-3000 Model 1456 CMOS image sensor operates immersed in LN$S$7, including cold start, and states in the conclusion that one CMOS image sensor was found to operate flawlessly immersed in LN$S$8 at 77 K (Graaf et al., 2023).

6. Applications, advantages, limitations, and terminological disambiguation

Rasnik’s original and most established application is detector alignment in high-energy physics, but the literature extends it well beyond that domain. The 2021 study lists alignment and deformation monitoring, seismic sensing / accelerometers, large-area sensor networks, and even long-baseline lunar deformation concepts. The 2025 study adds precision displacement sensing in VATIGrav, dynamic characterization with a Watt’s linkage, and thermal-vacuum qualification of LISA Quadrant Photoreceiver hardware. The C$S$9 study extends the concept to cryogenic accelerator metrology (Graaf et al., 2021, Graaf et al., 2023, Koushik et al., 3 Oct 2025).

Several advantages recur across these domains. Rasnik provides direct optical readout of 2D displacement, sensitivity to multiple degrees of freedom, and absolute coding with no cumulative drift. Its operating span is reported from $\theta_z$0 to $\theta_z$1. The 2025 formulation emphasizes electromagnetic immunity through a purely optical measurement principle, while the C$\theta_z$2 paper contrasts Rasnik with stretched-wire wire position sensor (WPS) approaches by noting the absence of broken-wire risk, wire sag uncertainty, and LN$\theta_z$3-flow-induced wire motion (Graaf et al., 2023, Koushik et al., 3 Oct 2025).

The limitations are equally specific. Rasnik is strongest in transverse sensing; longitudinal displacement is inferred indirectly through magnification and receives less detailed characterization. FFT-based reconstruction introduces a periodic linearity error that can dominate the shot-noise floor in dynamic use. In chained systems, uncertainty grows approximately as $\theta_z$4, so long alignment networks require extra anchors or multiple parallel chains. In cryogenic deployment, optical performance may be excellent, but overall accuracy is dominated by mechanical interface and calibration offsets of order $\theta_z$5 per Rasnik data point in the C$\theta_z$6 estimate (Graaf et al., 2023, Koushik et al., 3 Oct 2025).

A final point of clarification concerns terminology. Rasnik is unrelated to Ras in molecular oncology. The cell-mechanics study “Collective stresses drive competition between monolayers of normal and Ras-transformed cells” examines HEK-GFP and HEK-Ras-mCherry monolayers differing by oncogenic H-RAS$\theta_z$7 expression and concludes that the transformed monolayer displaces the wild-type monolayer through larger collective front stresses. That work concerns Ras-driven tissue mechanics, not the Rasnik optical alignment system (Moitrier et al., 2018).