Controls¶
The control stack, the dose-delivery timing, the sample-custody seam, and the offline-readout seam. First cut; handles read from the profile collection, carried confirm.
XFP runs on the NSLS-II EPICS / ophyd control stack, the same floor as the rest of the NSLS-II fleet, with a handful of PV-bound fluidic actuators on top. CORA observes that floor and, where it replaces bluesky-style orchestration, conducts over it; it does not replace EPICS itself.
Device handles¶
The control handles are filled from the beamline's own bluesky profile collection (NSLS2/xfp-profile-collection, the startup/ files), so the descriptor carries the real PV roots. The optics zone carries the bendable mirror (XF:17BM-OP{Mir:1}, OPT-1) and the slits (FE:C17B-OP{Slt:1}, XF:17BMA-OP{Slt:ADC}, OPT-2); the dose gating carries the shutters (XF:17BMA-EPS{Sh:1}, XF:17BM-PPS{Sh:FE}) and the delay-generator dose timer (XF:17BMA-ES:2{DG:1}, DOSE-1); the endstation carries the sample stages (XF:17BMA-ES:1{Stg:5}, XF:17BMA-ES:2{Stg:7}), the delivery pump (XF:17BMA-ES:1{Pmp:02}, FLOW-1), and the flux monitors (XF:17BM-BI{EM:1}, DET-1). They remain confirm-pending: a value read from the profile collection is evidence to verify with staff, not a CORA-owned fact (CTRL-1).
The dose-delivery timing¶
The dose is the experiment, so its timing is worth spelling out. There are three ways a dose is delivered, and CORA conducts each over the ControlPort:
| Dose mode | How the exposure is timed | Devices conducted |
|---|---|---|
| Seconds-scale | software-timed: open the pre-shutter, wait the exposure time, close it | DoseShutter (Shutter), FluxMonitor |
| Millisecond | the DG535 delay generator sets an opening-time and fires the Uniblitz fast shutter once | DoseTimer (TimingController), the Uniblitz (a PV-less shutter downstream of the timer), FluxMonitor |
| Shutterless high-throughput | the HTFly stage sweeps through the defining slit at a set velocity; exposure = slit gap / velocity | HtFlyStage (LinearStage), DefiningSlit (Slit), FluxMonitor |
In every mode the FilterWheel sets the dose rate and the FluxMonitor records the flux time-series; the delivered dose is exposure time times flux times attenuation. The dose timer (DoseTimer) binds TimingController because its anatomy is a settable opening-time plus a single-shot trigger, the same timing-box role the fleet's other position-capture / trigger boxes fill; the Uniblitz it fires has no EPICS PV of its own and is modelled downstream of the timer (DOSE-1).
The sample-custody and offline-readout seam¶
XFP's run produces a footprinted sample and a dose record, and the structural readout is offline. Three pieces sit in the seam rather than the device tree:
- The fraction collector (
XF:17BM-ES:1{FC:1}) is a PV-bound aliquot-routing actuator (a collect / waste valve, a tube index, a fill pattern, home). No existing Family fits an aliquot-router cleanly, and at n=1 CORA does not coin aFractionCollectorFamily; it is carried in the sample-custody seam, the hand-off that captures footprinted aliquots into tubes for offline analysis (FC-1,READOUT-1). - The 96-well plate is addressed in pure Python (8 columns x 12 rows, a coordinate table, no robot and no PV); moving to a well is a move on the
HighThroughputStage. It is a Procedure over the spine plus a Subject custody thread, the i03 / MX3 / LIX custody-as-Procedure precedent, here at the no-robot end of that spectrum (HT-1). - The offline mass-spectrometry readout is downstream, off the beamline, and absent from the profile collection. CORA keeps the dose record as the system of record for a footprinting run; the MS structural analysis is a separate, later step (
READOUT-1).
The orchestration seam¶
The XFP acquisition runs through bluesky plans (gate a timed dose, or sweep the HTFly stage, while the pump flows the sample and the QuadEM records the flux trace), publishing documents to Kafka with run metadata in Redis; there is no Tiled and no queue-server in the profile collection (CTRL-1). That orchestration is the seam CORA's edge replaces: CORA conducts the dose program over the ControlPort, driving through ophyd / EPICS and the fluidic actuators rather than replacing them. The temperature / bias diagnostics (an SR630 thermocouple monitor, the Sydor bias controller) are read-only alignment-flux proxies, deferred (TEMP-1).
Equipment protection¶
Only the front-end photon-shutter enable status (XF:17BM-PPS{Sh:FE} enabled-status, interlock-derived) is in the profile collection; plans refuse to open it when disabled. The rest of the PSS search-and-secure permit signals are absent and not invented here (PSS-1). Because the experiment is high-flux white-beam irradiation of biological samples, the dose itself is a hazard the safety tier would gate; that mapping is not modelled in this cut and is carried against the safety questions (PSS-1, DOSE-1).