Master-Thesis-2024

Selected experimental results

This folder contains compact result records with their comparison windows and interpretation stated explicitly. It is not a raw-data release.

Experiment records

Record	What it contributes
Exp-H long-run OPL + pixel benchmark	Temperature-only feedback under a strong pressure disturbance; establishes the residual limit that motivated a fine correction layer.
March hybrid TEC-AO recovery experiment	44.36-hour disturbed run; documents sub-pixel final recovery together with the full-run limitation.
Why perfect zero-pixel regulation was not yet achieved	Control-report explanation of why the system differs from a textbook step-response problem, supported by Stage-2 metrics and a limitations map.
June 2026 local thermal-impulse repeat campaign	Four-location component-sensitivity study comparing grating and camera-mount thermal impulses with OPL, centroid and environmental context.

Cross-experiment indicators

Area	Reported result	Interpretation
Environmental optical-path response	approximately 0.4 micrometres per hPa; approximately 6 micrometres per degree C	Pressure and temperature both produced measurable optical-path-length variation.
Thermal control	standard deviation of temperature, sigma T = 2.47 mK over 44.5 h; 17.4 mK peak-to-peak	The TEC maintained millikelvin-level stability during a long closed-loop run despite an external temperature change of approximately 3.2 degrees C.
Active-optics calibration	approximately 0.15 px per step	The active-optics unit had sufficient authority for small centroid corrections, but its finite travel requires range management.
Fine correction test	mean residual 0.443 px to 0.153 px across 8 automated corrections	AO reduced small residual centroid offsets after calibration.
Hybrid control comparison	dY RMS 2.72 px to 0.69 px	The selected comparison interval showed a 74.5% reduction when thermal control was combined with AO fine trim.
Local thermal-impulse repeat	Grating A: strongest centroid response; Camera mount A: strongest OPL response	Establishes repeat priorities, not a final root-cause assignment.

Raw telemetry, laboratory configuration and detailed controller versions are not published here. The public record provides the measured scale of the problem, the control architecture and a hardware-independent decision model.