Download presentation
Presentation is loading. Please wait.
1
Hardware Injection Status Update
Eric Thrane, Adam Mullavey, Peter Shawhan, Jeff Kissel, and David Shoemaker on behalf of the HWINJ Team
2
Overview Critical infrastructure is set up and mostly tested.
Hardware injections demonstrated in ER7 (or before): CW, CBC, Burst, Stochastic. Imperfect injection recovery: there appears to be a not-yet-understood sign-flip in H1 filters. Thank you to Technical Review Board (TRB) for helpful feedback; some recommendations implemented now, others saved for later due to limited FTEs.
3
Better Front-End Integration Now than in iLIGO
Front-End Calibration Layout Diagram,
4
ER7 Injection Software Injections carried out at H1 and L1 using version-controlled code. Transient injections carried out with tinj: a wrapper for awgstream. Request from TRB to migrate to Guardian in the future. CW injections carried out with psinject + makefakedata (LAL code). Stochastic injections carried out before ER7 with awgstream. HWINJ SVN: DOCUMENTATION:
5
ER7 Debugging A number of bugs identified and fixed. An illustrative (but not exhaustive list): EPICS flag identifying injection type (burst vs. CBC) was backward. tinj died every 12.5 hours due to book-keeping error with file identifiers. tinj modified to check inent bit CW injection of L1 data at H1 For details, see Peter Shawhan’s wiki page:
6
ER7 Logging & Segment Generation
ODC uses front-end time series data and EPICS channels set by HWINJ code to set bitmask bits SegGener makes segments to go into dqsegdb Significant debugging due to confusion from last-minute changes to ODC bits, but eventually ODC bits and segments shown to behave properly. Injection results logged by injection code, but… Injections not yet marked in GraceDB (GRiNCH). Injection logs not archived. [sysadmins working on] No automated cross-check of ODC bits, segdb, log files, schedule etc. [being developed by Maryland grad student] RED: requires assistance from DA groups. PURPLE: priority action item for HWINJ team…in progress
7
ER7 Injection Recovery Mostly successful recovery with CBC, Burst, CW…ran out of time for coherent stochastic injection. See Peter’s wiki page for details. All recoveries appear to indicate a flipped sign in the H1 inverse actuation function filter. Investigations ongoing: focus is on construction of a more suitable filter for O1. Also significant calibration error apparent:
8
Recovered Burst Injections (Shawhan) -- Basically consistent with what Matt Pitkin found from CW injections Up to ~80% amplitude error phase delay + sign flip
9
Current Limitations of Injection System
Saturation of injections at high frequency due to filter LIMIT LIMIT setting could be relaxed, but how much? Hard saturation will occur in the chain somewhere Limited implementation of operator override controls Basic enable/disable works; is checked 5 minutes before scheduled injection time No automatic restart of injection software Keith T. currently working to be able to run programs under monit
10
Blind Injections No blind injections performed during ER7.
Blind injection code (simple!) added to version control since ER7. The blind injection team is in the process of beta testing the blind injection software now. We expect to be able to demonstrate blind injections before and during ER8.
11
Next steps Working to address most pressing improvements requested by TRB, especially archival of injection logs. Test blind injection infrastructure. New inverse actuation filter. Test recovery of burst/CBC/CW/stochastic injections with revised code during ER8. Test segment generation during ER8.
Similar presentations
© 2025 SlidePlayer.com. Inc.
All rights reserved.