Comments from the 07/14/05 LCLS SLC-Aware Design Review: 
	Beam Synchronous Acquisition and Controls

Presenters: Diane, Debbie
Attendees:  Patrick, Till, Dayle, Kristi, Stephanie, Stephen, Nancy, 
	    Mike Zelazny, Ron Chestnut, Ron Mackenzie, Greg White, Bob Fuller
	    
Specification:
http://www.slac.stanford.edu/grp/lcls/controls/global/sw/slc_ioc/bsac/BSACDesign.htm	

Vdct drawings:
(1)Markup from hardcopies (Q=1 at top of MEASCNT, PUIDCNT, LOCA = slc ioc 
   location name,  userid = user name typed into EDM for CA)
(2)Change name of CheckPNET to CheckEVR (there can be 8 extra bytes for 
   LCLS purposes aside from epicsTS)
(3)Make new .db file with dirty sine wave test signal and controls for 
   diagnostic test on/off (future)
(4)Add FLNK to compress records that buffer unit data for CA clients / md 
   (future)
(5)Make a note that the drive-high (DRVH) limit for 
   MDEF:$(LOCA):$(MD):MEASCNTMAX needs to be at least 2800 but can be made 
   larger later if needed. 
(6)Document the upper limit on AVGCNTMAX.
(7)In order to display the time when the meas def is setup, please add a 
   timestamp record FLNKed from MD CTRL - it'll just be used to display the 
   time that the meas def was activated or deactivated.
(8)For bits beyond PNET that Patrick will be using, add in inclusion and 
   exclusion masks 4 and 5 to the MDEF database and include them as inputs to 
   CheckEVR.  bsacHdlr just sets them to 0.


Design doc:
(1) Reformat to make printable
(2) Insert updated BPM Overall.vsd
(3) Note that devices other than BPMS can have on-line/off-line/maint/checkout 
    states in epics.
(4) Design/doc collection structure in BSACLrgBuffer.vsd using array indices
(5) We could use a CA monitor for EPICS changes (or is it easier code-wise to 
    keep orig bsacAsync design)?  Also, need to keep the previous values of 
    the applicable SLC database secondaries for comparison purposes.  Need to 
    log an informational message when the SLC database is changed due to an 
    EPICS change.  Document that the message is not needed for the EPICS 
    change since ChannelWatcher will do the message.
(6) Look into semaphore necessity around measdef timeout R/Ws
(7) Set SCAN = Intr_IO last in sequence for measdef setup
(8) Note that there is no db consistency checks between EPICS and SLC db upon 
    initialization; immediate reply messages to SCP will complain if there 
    are any mismatches
(9) Note that SCP max nrpos = 2800; max navg = ?; we can change, if needed
(10)EPICS Device Supports do a scan IO request. (Decouples; better than FLNK 
    and having to process disabled records) (Our GO$MDs are automatically 
    managed and placed into list) to cause GO$MD records whose 
    SCAN = IO_Intr to process.  Until the device support is ready, plan on 
    FLNKing fanout records from the last ai/puid record to the 20 GO records 
    and code bsacHdlr to set the GO DISV to 0 (to stop) or 1 (to start) 
    instead of SCAN to IO_intr or Passive.
(11)BPMS x, y, tmit must be processed prior to the scan IO request.
(12)Pulse id, timestamp must be ready 1st before all other primaries.
(13)Draw a timeline of when EVR, BPMS, other BSAC devices, FFback need to 
    process relative to one another and the timing system (Steph).
(14)Abort cases: bsacAsync timeout, bsacHdlr setting abort flag due to 
    mismatched counters/EPICS error reporting in message, CLR_SCP command, 
    MEAS_STOP command, user sets via CA: MEASCNTMAX = 0,   any others?
(15)Double check abort logic
(16)No qraw is being reported for any devices. This would not have meaning to 
    maintenance personnel anyway because it is not raw camac, (raw VME), 
    which is not what they're used to seeing. They must go to device 
    (ie Till's) EDM Diagnostic display instead
(17)Check MDEF:USERID to see if measdef is brand new or if it replaces this 
    scp's current measdef.		
(18)Document that we assume one location name per IOC.  That IOCs cannot span 
    multiple locations (ie, one IOC cannot span both the injector and LINAC 
    sector 21).
(19)Document that the task that processes your 120hz records must run at a 
    high priority.
(20)Document that the PUID processing will be 360hz (the PP match will fail 
    for the unused time slots).
(21)Add diagnostic fields to bsacMD_ts structure.  Add iocsh routines that 
    dump bsacMD_ts and any interesting internal arrays/linked lists.
(22)For the bsacHdlr normal processing loop description, make sure the 
    immediate reply back to the SCP is done only after the large reply buffer 
    is allocated, the pointers and counters are set up, and all dbPut 
    operations are complete.  The SCP needs to know if there is any error on 
    any of these operations and the measurement never started.
(23)The "abort flag" in the BSAC_DATA and BSAC_DONE message processing detects 
    nrpos inconsistencies (like too many messages reported/unit). An error can 
    be logged for that unit in that processing.  MEASCNTMAX be set to 0 
    for an abort.
(24)For BSAC_DATA, BSAC_DONE, and BSAC_ABORT, also update diagnostics kept in 
    bsacMD_ts.
(25)Move the last bullet under BSAC_DONE processing to BPMO_GETREMDAT 
    processing.
(26)If any message to the bsacHdlr is missed, the measurement will eventually
    timeout.  There will be an overall counter/measDef  that will tabulate 
    DONE messages for units; it's max = #units * nrpos. This will quickly 
    tell us, during the DONE processing, when our measDef has completed the 
    cycle.
(27)For XTOL, ANGL, and OFFS, don't do any coding for them in bsacAsync - 
    make it all TBD.
(28)Add ability to simulate BPM data.  Don't need Toroid etc.
    X,Y, and TMIT are needed, since high level apps will assume the data 
    is bad unless tmit is reasonable. Other than non-0 tmit the fake data 
    need not be especially believable - a sine function with some error 
    would be ideal.  Using the record simulation features will probably 
    work.  The simulation can be turned on/off without rebooting the IOC and 
    still be driven by Till's triggers and use the incoming EVR patterns.