Multi-beam Optical Seeing Sensor (MOSS)

• MOSS is a direct, line-of-sight sensor that quantifies local optical-path turbulence in telescope enclosures using four near-parallel laser beams.
• It leverages differential centroid motion of focal-plane spots to assess dome and mirror seeing, aiding in thermal and airflow management.
• Prototype tests on AuxTel demonstrated its capacity to track turbulence variations influenced by temperature gradients, wind, and enclosure geometry.

Multi-beam Optical Seeing Sensor (MOSS) is a direct local-seeing monitor designed to measure optical-path turbulence inside and near a telescope enclosure rather than the free-atmosphere seeing conventionally characterized by DIMM, MASS, or related site monitors. In the reported implementation, MOSS outputs four near-parallel beams of light that travel along the optical path and are imaged by the telescope detector, landing like starlight on the focal plane; the differential motion of the resulting spot images is then used to characterize dome and mirror seeing along the actual telescope line of sight (Marangola et al., 1 Aug 2025).

1. Scope, definition, and observational niche

MOSS targets the component of image degradation produced by refractive-index inhomogeneities in the air through which starlight passes inside the dome and immediately around the telescope structure and optics. The instrument is motivated by the observation that local turbulence inside an enclosure is shaped by boundary conditions, enclosure geometry, airflow, thermal gradients, mirror thermal mismatch, slit orientation, and wind coupling, so standard upper-atmosphere seeing formalisms do not straightforwardly recover it from conventional telemetry (Marangola et al., 1 Aug 2025).

The defining feature of MOSS is therefore not merely that it is multi-beam, but that it is a direct line-of-sight sensor for below-the-slit optical degradation. The operational premise is that the dome environment is more controllable than the upper atmosphere. Measurements of local optical-path turbulence can consequently inform thermal control, airflow management, dome operation, and mirror conditioning. This places MOSS in a different category from site monitors whose principal outputs are integrated atmospheric parameters such as seeing angle, $r_0$, $\tau_0$, or $C_n^2(h)$ profiles (Marangola et al., 1 Aug 2025).

The published prototype paper also makes clear what MOSS is not. It does not present a full analytical turbulence model for enclosure seeing, and it does not provide a DIMM-like or Shack-Hartmann-derived inversion from differential image motion to canonical atmospheric parameters. Instead, the central observable is the standard deviation of pairwise PSF separation as a function of beam separation, interpreted as a measure of local optical-path turbulence and its spatial coherence (Marangola et al., 1 Aug 2025).

2. Instrument architecture and optical principle

The reported MOSS prototype was installed on the 1.2 meter Auxiliary Telescope (AuxTel) at the Vera C. Rubin Observatory in Chile. It uses a Quarton VLM-520-61 LPO low-divergence laser diode, with wavelength $515\text{–}530\,\mathrm{nm}$, output power $< 0.39\,\mathrm{mW}$, and half-angle divergence $< 0.4\,\mathrm{mrad}$. Beam expansion is provided by connected Thorlabs BE20-532 and BE05-532 beam expanders, attenuation by an OD 2.5 ND filter on a Thorlabs MFF102 flip mount, and beam splitting and steering by Thorlabs BS031 beamsplitter cubes and BB2-E02 mirrors. Remote pointing control uses Thorlabs APY002 tip/tilt stages with PIA13 piezo actuators; control electronics use a LabJack for laser power and control and an Intel NUC for remote operations (Marangola et al., 1 Aug 2025).

Aspect	Value	Function
Light source	Quarton VLM-520-61 LPO, $515\text{–}530\,\mathrm{nm}$, $<0.39\,\mathrm{mW}$	Produces strobed artificial-star beams
Beam train	BE20-532, BE05-532, OD 2.5 ND, BS031, BB2-E02	Expansion, attenuation, splitting, steering
Pointing and control	APY002 with PIA13, LabJack, Intel NUC	Remote alignment and operations
Beam geometry	Four beams, mounted 15.24 cm apart on the breadboard	Samples multiple near-parallel optical paths
Imaging cadence	2 s exposure, 2 s readout	Records focal-plane spot motions

The optical train divides one laser beam into four beams of roughly equal flux and aligns them so that they propagate along the telescope boresight and strike the telescope pupil along a chord of the primary mirror. The beams are nearly parallel so that the telescope treats them like light from distant point sources. The paper notes that parallel light beams that hit the primary mirror converge to the same point on the focal plane, so the beams are made slightly off-parallel in order to form four distinct focal-plane spots (Marangola et al., 1 Aug 2025).

After beam expansion, the beam diameter is approximately 50 mm, yielding diffraction-limited PSFs of about 3.5 arcsec FWHM on AuxTel. In the actual processed data, the measured post-convolution beam FWHM values were 6.59 arcsec, 5.89 arcsec, 5.47 arcsec, and 8.86 arcsec for beams 1 through 4. The spots are not perfectly Gaussian because of the laser diode, but each beam’s PSF shape is stable from image to image, which is important for reliable centroiding (Marangola et al., 1 Aug 2025).

Temporal freezing is implemented by strobing the laser. The paper states that a strobed light source can “freeze” the instantaneous index variations transverse to the optical path. Minimum laser pulses possible were 10 $\mu$s; tested pulse lengths included 100 ms, 0.1 ms, and 10 $\mu$s. The reported best compromise was 0.1 ms: 100 ms pulses averaged over the turbulence and suppressed differential image motion, while 10 $\tau_0$0s pulses were thought to suffer from laser-diode thermal turn-on transients. With stronger attenuation, adjusted remotely to OD 3.5, 1 ms pulses are also possible (Marangola et al., 1 Aug 2025).

3. Observables, centroiding, and pairwise differential motion

The primary MOSS observable is differential centroid motion, not absolute image position. For each exposure, centroid positions are measured for all four focal-plane spots, and the six pairwise separations are then calculated. The analysis uses the standard deviation of these pairwise PSF separations over a stack of images. Because the metric is based on pair separations rather than absolute positions, substantial common-mode motion from wind-induced dome motion, slight telescope motion, or other global disturbances is removed at the measurement-definition level (Marangola et al., 1 Aug 2025).

The reduction sequence begins with a raw CCD image containing four laser spots. The beam region is convolved with a Gaussian kernel of FWHM 35.32 pixels, after which centroids are computed with the LSST Science Pipelines. This convolution is used because the raw beam profiles are non-Gaussian due to the laser diode and smoothing improves centroiding stability. Beam identity on the detector was established by physically blocking each beam and taking an image, since the focal-plane arrangement does not necessarily mirror the optical-bench ordering (Marangola et al., 1 Aug 2025).

For each beam pair, separations in $\tau_0$1 and $\tau_0$2 are tracked frame by frame. A running mean with a window of 20 points is subtracted to remove slow drift, and the standard deviation of the residual separation is then computed. These standard deviations are plotted against the separation of the corresponding beams on the pupil. The interpretive logic is explicit: nearby beams should traverse more similar refractive structures and exhibit smaller differential motion, while more widely separated beams should decorrelate and display larger differential motion; a plateau would mark the scale at which the motions become uncorrelated (Marangola et al., 1 Aug 2025).

The only explicit equation given for measurement precision is the centroiding uncertainty,

$\tau_0$3

with reported beam-by-beam values of $\tau_0$4, $\tau_0$5, $\tau_0$6, and $\tau_0$7. The average centroiding uncertainty is $\tau_0$8, and the uncertainty in a separation value is given as $\tau_0$9. The paper states that these uncertainties are far below the measured differential motions, so detector and centroid noise do not dominate the results (Marangola et al., 1 Aug 2025).

Equally important is what the paper does not formalize. It does not write an explicit differential-image-motion variance law of DIMM type, does not provide a structure-function model for enclosure turbulence, and does not give a direct conversion from pairwise separation statistics to $C_n^2(h)$0, $C_n^2(h)$1, or a conventional seeing FWHM. The reported interpretation is instead that the standard deviation of PSF separation “maps to the diameter of the dome and mirror seeing disc,” with the published numerical result expressed as a lower bound rather than a fully inverted atmospheric parameter (Marangola et al., 1 Aug 2025).

4. Prototype deployment on AuxTel and reported results

The prototype was mounted on the dome at approximately $C_n^2(h)$2 elevation, adjacent to the dome slit to minimize telescope repointing required for MOSS exposures. The hardware was mounted on Thorlabs MB648 breadboards, with support hardware attached to horizontal dome structures using hose clamps; the main board was supported with a right-angle bracket and turnbuckles. Beam pointing was remotely adjustable over a total range of $C_n^2(h)$3 (Marangola et al., 1 Aug 2025).

The AuxTel camera is described as a $C_n^2(h)$4 deep-depletion CCD with 16 readout amplifiers, using 2 s exposure time and 2 s readout time, so frames are separated by approximately 4 s wall-clock cadence. The paper states that a series of 50 images takes less than five minutes, allowing interleaving of MOSS sequences with normal astronomical observing. For the main closed-dome analysis, the principal sequence contained 1000 frames, of which 878 passed centroiding quality cuts (Marangola et al., 1 Aug 2025).

The main reported result is a lower bound on optical path turbulence of $C_n^2(h)$5. In the closed-dome dataset, the standard deviation of differential PSF motion increased with beam separation, with significantly larger motion in $C_n^2(h)$6 than in $C_n^2(h)$7. The $C_n^2(h)$8-direction points showed flattening, suggesting that the correlation scale had been approached in that direction, while the $C_n^2(h)$9-direction points continued to rise, indicating that the characteristic scale had not yet been reached vertically. The paper interprets the larger $515\text{–}530\,\mathrm{nm}$0-direction signal as stronger vertical temperature fluctuations in the dome (Marangola et al., 1 Aug 2025).

With the dome open, MOSS was used to examine nightly evolution. The reported conditions included a particularly windy night with gusts up to 17 m/s; later discussion also notes gusts exceeding 15 m/s. Under these conditions, optical-path turbulence increased above the $515\text{–}530\,\mathrm{nm}$1 lower-bound level seen in the closed-dome dataset, the differential-motion curves became flatter with separation, and the difference between vertical and horizontal seeing became less significant. The authors interpret this as wind reducing the vertical temperature gradient while increasing the overall optical-path turbulence, and they note that the turbulence increased significantly when wind speed spiked (Marangola et al., 1 Aug 2025).

Uncertainties were quantified with 67% confidence intervals computed via bootstrapping the pairwise separations from the image stack. The paper also notes that different beam pairs with the same separation can yield different standard deviations, implying that the turbulence is not translation invariant. It further suggests, from smaller-subset analysis, that the coherence time in $515\text{–}530\,\mathrm{nm}$2 appears shorter than the full 1000-image set, and concludes that the optimal image-sequence length and cadence remain to be determined (Marangola et al., 1 Aug 2025).

5. Relationship to other multi-beam and focal-plane seeing sensors

MOSS belongs to a broader family of instruments that use simultaneous optical channels, multiple subapertures, or multiple beam paths to characterize delivered image quality, but its sensing target is more local and more enclosure-specific than most. Active-optics Shack-Hartmann seeing work showed that long-exposure subaperture spot sizes at the telescope focus can provide real-time seeing estimates when treated with an OTF-based algorithm, with finite-$515\text{–}530\,\mathrm{nm}$3, wavelength, sampling, and detector-PSF corrections explicitly modeled (Martinez et al., 2012). A subsequent dual-arm, simultaneous multiwavelength AOSH concept extended this logic to joint estimation of seeing and turbulence outer scale from wavelength-dependent spot FWHM, emphasizing measurement at the telescope focus rather than at an external monitor location (Martinez, 2014).

A more explicitly multi-beam night-time monitor is SHIMM, which replaces the two-hole DIMM mask with a Shack-Hartmann wavefront sensor and uses many subapertures to estimate seeing, a low-resolution three-layer turbulence profile, and—at higher frame rate—an atmospheric coherence timescale. In that architecture, seeing comes from centroid auto-covariance fitting, while scintillation observables supply coarse altitude discrimination (Perera et al., 2023). By contrast, WFWFS/S-DIMM+ uses many subapertures and many field angles on the Sun to form covariance maps of differential image motion and invert them into a layered turbulence profile, making it a multi-directional differential-tilt profiler rather than a local enclosure-turbulence sensor (Tham, 2011).

Within interferometric instrumentation, the GRAVITY acquisition and guiding camera is especially close to a MOSS-like concept in the sense that it is a simultaneous four-beam sensor of the state of delivered telescope beams. It measures field tip-tilts, lateral and longitudinal pupil shifts, pupil images, and quasi-static aberrations for four beams in parallel, but its purpose is beam stabilization for interferometric sensitivity and astrometric precision rather than canonical seeing estimation (Anugu et al., 2018). A different hardware lineage is represented by miniaturized Starbug-compatible Shack-Hartmann sensors, which show that many deployable wavefront-sensing probes can be multiplexed onto one detector and explicitly identify “seeing monitors” among their possible applications (Goodwin et al., 2014).

Against that background, MOSS is most precisely described as a direct, line-of-sight, local optical-path differential image-motion sensor. Unlike the AOSH-derived methods, it does not explicitly recover $515\text{–}530\,\mathrm{nm}$4 or $515\text{–}530\,\mathrm{nm}$5. Unlike SHIMM or WFWFS/S-DIMM+, it does not reconstruct a turbulence profile. Unlike GRAVITY’s beam-state sensor, it is not primarily an interferometric guiding subsystem. Its distinctive niche is the direct measurement of dome and mirror seeing along the science path inside the enclosure (Marangola et al., 1 Aug 2025).

6. Interpretation, limitations, and future development

Several limitations are explicit in the published MOSS work. First, the instrument measures the combined local contribution of dome, mirror, and other below-the-slit optical-path turbulence; it does not independently decompose these components. Second, spatial sampling is limited to four beams, giving six baselines and only chord-like coverage across the pupil. Third, the absence of translation invariance in the data complicates any simple isotropic structure-function interpretation, since beam pairs with the same separation can behave differently (Marangola et al., 1 Aug 2025).

A further limitation concerns calibration to standard seeing metrics. The paper does not provide an explicit algebraic conversion from the measured separation-statistics curve to the quoted lower bound of $515\text{–}530\,\mathrm{nm}$6, and it does not specify whether that bound is set by the largest measured pairwise standard deviation, a quadrature combination of $515\text{–}530\,\mathrm{nm}$7 and $515\text{–}530\,\mathrm{nm}$8, or another estimator. The result is therefore best read exactly as presented: a lower bound on local optical-path turbulence, not a fully inverted DIMM-like seeing estimate (Marangola et al., 1 Aug 2025).

This distinction matters because a common misconception is to treat MOSS as a replacement for standard atmospheric monitors. The published work does not support that interpretation. The paper explicitly contrasts MOSS with DIMM, MASS, SLODAR, and SCIDAR-type systems, noting that those methods are designed for upper-atmosphere seeing or altitude profiling, whereas MOSS directly senses the optical consequence of local enclosure turbulence along the boresight. It also notes that MOSS was strongest, in its initial publication, as an instrumentation and proof-of-concept demonstration rather than as a complete turbulence-inversion framework (Marangola et al., 1 Aug 2025).

The stated future directions are correspondingly practical. They include determining the optimal number of images per MOSS sequence, determining optimal sequence frequency during nighttime operations, exploring temporal coherence with different pulse lengths, combining MOSS with anemometer and thermal telemetry, increasing the number of beams and extending their separations, and deploying similar systems on larger telescopes, including the Rubin Simonyi Survey Telescope (Marangola et al., 1 Aug 2025). A plausible implication is that later MOSS-like systems may evolve toward tighter integration with observatory control loops, but the published prototype establishes the concept primarily as a direct monitor of local optical-path turbulence where operational intervention is possible.