J. Kissel, R. Bork, D. Barker, J. Batch
After installing and completing all of the changes to the front-end simulink model for the QUAD, it took a while before it would successfully compile, due to my own misunderstandings of how tags and (separately) cdsBinIns work. Unfortunately both bugs result in a bunch of red-herring errors being thrown. Many thanks to Rolf, Jim, and Dave for their help resolving these issues*. Once I got that up and running, the rest of the day was spent updating the QUAD OVERVIEW screen to match the new infrastructure. This is almost entirely complete, but I have not made any of the necessary changes to the sub screens. Tomorrow is another day.
For the record, CDS would like to begin updating to RCG 2.8 tomorrow. I consider all of the QUAD models ready for this upgrade, though it would be nice to see the remaining QUADs (see below) installed against RCG 2.7.2, before going forward. The remaining suspension types are *not* ready, and should be left alone.
I have left the QUADs as follows:
ETMY -- compiled, installed, restarted, restored (but the safe.snap will need to be retaken because of all the new bits).
ETMX -- compiled, but not installed
ITMY -- compiled, but not installed
ITMX -- compiled, but not installed
Robert and Arnaud are taking measurements on the other three tonight, and need a good DAQ status, so I have refrained from installing new stuff.
Note -- the good news is that with the removal of the ISI OFFLOAD path and reduction of the DRIVEALIGN matrix, the QUAD models now finish all computation in 37 [usec] as opposed to 45 [usec] of their 61 [usec] clock cycle.
Note -- Stefan's new dependency script is also awesome (see LHO aLOG 8213).
Note -- The compilation process only spits out he last 10 errors or so. To get a full list of errors, look at ${BuildLocation}/${modelname}.log on the build machine. At LHO, for the ETMY, that's /opt/rtcds/lho/h1/release/h1susetmy.log
Note -- For a complete and utter gorl details of what the parser spits out -- and therefore a detailed list of every part's inputs and outputs, look at ${BuildLocation}/src/epics/util/diags.txt after a compile. It's daunting, but if you have one channel or two that's problematic, you can just search for it and investigate it's connections to see if they make sense.
Note -- cdsBinIns MUST have their input grounded. Otherwise, the RCG claims that everything connected to them -- all the way out to the top-level model -- is disconnected, instead of telling you what's really wrong. Rolf says this bug is on the list of things to be fixed when he goes through all parts and makes sure certain ones don't require things attached to them for the model to compile.
Note -- Within a given sub-system level, you cannot pipe an input into a sender tag, then stick the output of the receiver tag directly into a sub-sub-system. To the RCG, this is like "skipping a level," because nothing happens to the signal at the sub-system level, so connections are lost. Similarly, when you do this, the RCG throws errors on *everything* connected to this path, but not tell you what's really wrong.
The new list of things that I've done to the front end code:
- Add SVN $Id$ to all models and library parts, to better track version control
- Moving the DRIVEALIGN matrix out of the DAMP/TEST/OPTICALIGN paths on the main chain, such that it only operates on the LOCK path for each stage (and therefore reducing the size of the M0 / M1 version)
- Adding output switches after the internal SUS offloading, so -- in the "offloaded" hierarchy scheme -- one can send alignment signals straight to the top without zeroing out diagonal DRIVEALIGN matrix elements
- Removing the wasteful, unnecessary DRIVEALIGN path on reaction chain's TOP mass.
- Making both DAMP and LOCK filter banks use cdsCtrlFilt, such that they produce a status bit word for ODC consumption during lock acquisition
- Defining the list of channels that shall stay in the science frames, to be stored for all time
- Removing obsolete L1 and L2 (lower stage) damping from the QUADs which had been needed for very early H2OAT operation before ISIs were turned on
- Removing the explicit definition of 2048 from the stored ODC DAQ channel, such that it is by default sored at the model's native rate as designed
- Adding a length-driven, (optical lever/OSEM) angle response lock-in for force-to-angle decoupling and tuning.
The remaining to-do list for the QUAD screens:
- Find the generic ISC Lock-In screen. They must have one.
- Add "GOOD"/"BAD" states to DAMP and LOCK screens
- Remove DRIVEALIGN from RO screen
- Create LKIN2OSEM matrix screens
- Add PUM Analog WD to OVERVIEW screen.
- Add Coil Driver BIO status information to the OVERVIEW screen.
Here's the outstanding userapps SVN status on files that are pertaining to my work:
? common/medm/quad/SUS_CUST_QUAD_L1_LKIN2OSEM.adl
? common/medm/quad/SUS_CUST_QUAD_M0_CENTERING_TEST.adl
? common/medm/quad/SUS_CUST_QUAD_M0_LKIN2OSEM.adl
? common/medm/quad/EDITME.adl
? common/medm/quad/SUS_CUST_QUAD_L2_LKIN2OSEM.adl
? common/medm/quad/SUS_CUST_QUAD_L3_LKIN2ESD.adl
M common/medm/quad/SUS_CUST_QUAD_OVERVIEW.adl
MM common/models/HAUX_MASTER.mdl
MM common/models/HLTS_MASTER.mdl
MM common/models/SIXOSEM_F_STAGE_MASTER.mdl
MM common/models/BSFM_MASTER.mdl
MM common/models/MC_MASTER.mdl
M common/models/OMCS_MASTER.mdl
MM common/models/HSTS_MASTER.mdl
MM common/models/FOUROSEM_STAGE_MASTER.mdl
MM common/models/TMTS_MASTER.mdl
MM common/models/QUAD_MASTER.mdl
M h1/medm/h1mcall.adl
M h1/models/h1sustmsx.mdl
M h1/models/h1sustmsy.mdl
M h1/models/h1susprm.mdl
M h1/models/h1sussrm.mdl
M h1/models/h1suspr2.mdl
M h1/models/h1susbs.mdl
M h1/models/h1suspr3.mdl
M h1/models/h1sussr2.mdl
M h1/models/h1sussr3.mdl
M h1/models/h1susetmx.mdl
M h1/models/h1susomc.mdl
M h1/models/h1susetmy.mdl
M h1/models/h1susmc1.mdl
M h1/models/h1susitmx.mdl
M h1/models/h1susmc2.mdl
M h1/models/h1susmc3.mdl
M h1/models/h1susitmy.mdl
(The double "MM" status means that I've modified the model itself, AND the svn properties to make sure the SVN $Id$ is recognized.)
