Optical surveys: sensors for the upgrade Adrian Bevan, Karen Hayrapetyan, Camillia Messiouni 1
Overview ROOT is too slow to process large data sets from our smartscope; so have adopted an analysis procedure using Matlab. Fast Felxible Easy to use.... but requires some of us to learn a new tool 2 Mechanicals are flat Sensors are bowed; either convex or concave depending on vendor/process details
QA model Collect data in a grid with a finite spacing between measurement points. – Time: 15min for a scan in (x,y) steps of 500 um Plot and fit the data points with a plane. – Time: seconds (gave up on ROOT) Histogram the residuals about the mean defined by the plane; use spread as QA metric. – Acquisition with the OGP software and post processing by Matlab is quick enough that we _can_ assay all sensors passing through our lab before electrical testing. – i.e. can quickly pass/fail sensors out of QA bound of 200um bowing. 3
What about residuals for alignment? This is a demonstration – would like to repeat with a module. Use CMS model for alignment of their strip modules as a basis for study. Aim: demonstrate the reduction in residual spread relative to a model. 4 Flat sensor orientation plane dfn. Parabolic bowing term Linear bowing term ATLAS equivalent; only use the first 3 terms and fit for 4 sensors glued in an ideal module. See: NIM A (2011) (v)
Results with a mechanical sensor Fitting with a planar model um spread for residuals Residual (data - fit model) Number of points
Results with a failed electrical sensor (1) Fitting with a planar model 6 150um spread for residuals Residual (data - fit model) Number of points
Results with a failed electrical sensor (2) Fitting with a planar model 7 200um spread for residuals Residual (data - fit model) Number of points
Results with a mechanical sensor Fitting with the CMS model um spread for residuals Residual (data - fit model) Number of points
Results with a failed electrical sensor (1) Fitting with the CMS model 9 ~45um spread for residuals (c.f. 150 for planar model) Residual (data - fit model) Number of points
Results with a failed electrical sensor (2) Fitting with the CMS model um spread for residuals (good description of most points, but long tails) Residual (data - fit model) Number of points
Results with a mechanical sensor Fitting with cubic terms um spread for residuals Residual (data - fit model) Number of points
Results with a failed electrical sensor (1) Fitting with cubic terms 12 40um spread for residuals Residual (data - fit model) Number of points
Results with a failed electrical sensor (2) Fitting with cubic terms um spread for residuals Residual (data - fit model) Number of points
Overview of results Three fit models used: – Planar (linear) – Quadratic surface – Cubic surface – At least a factor of 3-5 improvement in the residuals measured for these sensors relative to a nominal planar test when applying a more sophisticated model. – Best case: is a x5-10 improvement for convex bow. – More statistics required to draw conclusive results. – We care about modules, not sensors, for the build; so need to repeat with modules. 14 Spread in Residuals Planar fitQuadratic fitCubic fit Mechanical Concave bow Convex bow
Local plank flatness (Stave Plank 7) Is the surface flat (if so how flat)? – on the scale of a module Did a laser surface scan over a region of the stave (100x100mm). Generally flat – can see undulations from tracks; LHS is the main power line track that sticks out of the surface (old design of the kapton bustape). Flat to better than +/- 50 um. Caveat: old tape design, new planks should be a lot better than this!
Summary Can reduce the hypothetical position error by a significant factor: – x3-5 with these sensor tests. – Want to repeat with modules to quantify how this would affect the real items going into the SCT / pixel system. Also plan to do cooling tests: exploring the possibility of having this as part of an ATLAS authorship task for a student. 16