A wobble visible in the data
The study describes a typical 50-Hz wobble in the SOAR telescope's optical axis, its line of sight, tracing a small ellipse and measuring 20 to 30 milliarcseconds (mas). In speckle observations, the vibration substantially reduced high-spatial-frequency signal and blurred the speckles: the analysis gives about 20 mas RMS motion and a 60-mas peak-to-peak image displacement. A 20-ms exposure sampled one full period of the 50-Hz vibration.
The result comes from an engineering case study of one 4.1-m SOAR telescope in Chile, not a participant sample. The work uses adaptive-optics and speckle records, accelerometer time series and mechanical excitation tests. Its scope is to characterize the vibration, examine likely sources and propagation paths, assess possible amplification by a fast M3 tip-tilt servo, and describe mount tracking jitter.
How the telescope was checked
For adaptive optics, the records used wavefront-sensor spot coordinates and deformable-mirror voltages. A typical record lasted 5 seconds and held more than 2,000 samples. The analysis split the data into four portions of 512 samples and averaged them to reduce spectral noise.
Speckle data cubes, sets of short-exposure images, typically contained 400 or 600 images, each acquired with a 25-ms exposure over runs lasting 10 to 15 seconds. Star centroids were saved for tracking analysis, allowing the study to follow image motion through each sequence.
The puzzle behind the signal
The accelerometer comparison made the source question harder. In the worst-case 50-Hz transverse acceleration measurement at M2, the signal was about 5 microvolts. The paper's solid-body calculation for 1 micrometre of motion produced 6.5 millivolts and 45 mas of angular amplitude. If those signals were connected, the analysis says strong amplification would be needed.
The author's working interpretation is that a mechanical perturbation propagated through the telescope and was apparently amplified by the fast M3 servo, although excitation within M3 could not be excluded. The source attribution remains open: the true source of the 50-Hz acceleration and the reason for its gradual disappearance after 2020 were not identified.
A separate 2018 test during the November coating shutdown tested the M3 optical response at 50 Hz. The result depended strongly on where the excitation was applied. At locations A through D, the X/Y RMS centroid motions were 58/269 mas, 62/114 mas, 60/236 mas and 89/455 mas. The response varied by location, but the test did not identify the vibration's unique source.
Several frequencies, several signatures
Other peaks pointed to several different mechanical signatures. The summary reports a 16-Hz resonance in the M2 unit and a 65-Hz resonance in the M1 support, with the 65-Hz component appearing in adaptive-optics astigmatism. It also reports a 47-Hz fan-related focus oscillation. The 47-Hz and 65-Hz effects were described as too small to affect delivered resolution. The report also contrasts a 20-Hz M2 resonance before LLT with a 16-Hz value after LLT and electronics were added.
Mechanical tapping tests gave another view of the M2 response. They recorded a 14.4-Hz X/Y resonance with a 1/e decrement of 6 to 10 seconds and a linear XY trajectory at 45 degrees relative to the spiders. A weaker 51.5-Hz resonance appeared when the LLT or vertical spider was tapped.
Tracking adds another layer
Mount tracking added a separate pattern of jitter. The analysis associated it with encoder position errors: encoder tape periods were 8.904 arcseconds and 20.24 arcseconds, while mount-servo bandwidth was about 1 Hz. Fast-tracking harmonics were amplified rather than compensated. In one example, elevation centroid motion was 0.27 arcseconds RMS against 0.12 arcseconds of encoder error, and azimuth motion was 0.08 arcseconds.
What the measurements leave open
The evidence has clear boundaries. The paper examines one telescope and does not report a total record count. The 20-to-30-mas wobble is described as typical rather than accompanied by a formal uncertainty interval. The study therefore provides an instrument-specific diagnosis, not evidence that the same pattern holds across other telescopes.
The same caution applies to the accelerometers. Their signals were not proportional to optical-axis motion, limiting their use as direct tilt proxies, and the study does not show that the accelerometers can directly sense or correct optical-axis tilt. The practical message is narrower: it maps several vibration signatures, shows that one can degrade speckle data, and leaves the 50-Hz source unresolved.
Paper data and sources
Original title: Study of Vibrations at SOAR
Authors: Andrei Tokovinin
Journal/Repository: arXiv
Status: Preprint, not yet peer-reviewed
First online: 2026-08-25
DOI: Not available
Original paper · Full text