Contains soft real-time software which runs on NI cRIO for M1M3 support. FPGA handling low-level communication, and LabView EUI (GUI panel) running on host PC are separate projects.
M1M3 Support System (M1M3SS) operates Dual Axis Actuators (DAAs) and Single Axis Actuators (SAAs) to keep mirror in neutral buoyancy state, counteracting primary gravity force on the mirror. Hardpoints forming hexapod are used to position the mirror. The control algorithm offloads force from top of the hardpoints into force actuators, so force reported by load cells placed on top of the hardpoints shall be close to 0 during normal mirror operations.
Force actuator target force is calculated as sum of the following components:
Forces calculated from X, Y and Z acceleration measured by 8 one axis accelerometers placed inside mirror cell (hence using the same reference system as the mirror cell). The force offsets are calculated using LUT, which assumes linear dependence of acceleration and resulting force. Input acceleration is in m^2/s.
Calculated and provided by AOS. Set with ApplyActiveOpticsForces, reset with ClearActiveOpticsForces. AOS process wave front sensing data, fits coefficient of the Zernike polynomials to produce Zernikes, and those are transformed into mirror deformation. The deformation is corrected by applying force offsets in Z direction. The force offsets (156 vector for 156 force actuators) are passed (in N) as parameter of the command. Forces resulting from transformation using the table are assumed to be in mN, so need to be divided by 1000.0 to get N forces.
Calculated from look up tables from telescope azimuth. Six values are stored into look-up table for X,Y and Z direction forces, representing five Five degree polynomial coefficients (thus enabling close approximation to goniometrical functions). The coefficient should be empirically fitted. Resulting forces are in N.
Balances hardpoint force excess. Takes in hexapod forces and moments calculated from excess forces, and produces force to be applied on force actuators. Resulting forces are in N.
Calculates gravity vector (elevation) compensation. As gravity is the major force acting on the mirror, this is the biggest force contributor. Similar to azimuth, the resulting force offsets are calculated using fifth degree polynomial. Fifth degree polynomial is enough to estimate sin and cos to more then 2 decimal points - and force distribution naturally resembles gravity vector projection on X and Y axis. Elevation is provided in degrees (0 on horizon, 90 in zenith). Resulting forces are in N.
Operator supplied offset forces. Operator specify offsets in X, Y and Z, and corresponding offsets moments. Those are distributed using the same transformation used for Balance forces. Resulting forces are in N.
Static forces to bend mirror close to optimal optic shape. Those are measured in a lab and stored in configuration tables. Resulting forces are in N.
Calculate correction forces for temperature. Fifth degree polynomial fit is assumed. Input is temperature in degrees centigrade, output is force in N.
Similar to acceleration forces. Cell velocity is provided by accelerometers placed inside mirror cell. Linear dependence between force and cell velocity is assumed.
DAAs and SAAs are also equipped with booster valves. Those are activated to make the actuator stiffer during mount accelerations.
Mirror is precise. Excessive force loads can break it. Control software must protect mirror from such forces.
The safest configuration for the mirror is when it's resting on supporting coils in passive state. Any failure or detected problem (seismic event,..) shall put mirror quickly to lowered position. Of course the controller and cell hardware shall not drop the mirror into a free fall. Emergency lowering of the mirror is referred as cell panic in the following text.
DAA and SAA are equipped with force cells. Difference between target and measured force is measured and excessive values leads to cell panic.
Hardpoints are equipped with pneumatic driven break-away mechanism. Any excessive force will make hardpoint break away, and once detected with control software, mirror goes into panic. Mirror is still safe with any number of loose hardpoints, resting on SAA and DAA.
Any miss-communication, unexpected behavior of either hardpoint, SAA or DAA shall bring mirror into panic.
NI cRIO is running FPGA code programmed in LabVIEW and C++ binary code. FPGA code handles details of low-level modbus communication. C++ code handles communication over SAL with TCS, and communicates with FPGA with a command FIFO
- see writeCommandFIFO and readCommandFIFO calls:
bool FirmwareUpdate::ClearFaults(int subnet, int address) {
this->desiredState = 3;
ModbusBuffer buffer;
this->SetupBuffer(&buffer, subnet);
buffer.writeU8((uint8_t)address);
buffer.writeU8(65);
buffer.writeU8(0);
buffer.writeU8(5);
buffer.writeCRC(4);
buffer.writeEndOfFrame();
buffer.writeWaitForRx(10000);
this->EndBuffer(&buffer);
this->fpga->writeCommandFIFO(buffer.getBuffer(), buffer.getLength(), 0);
return this->ProcessResponse(&buffer, subnet);
}The command is processed inside FPGA, put into command queue for serial/modbus port, send over the serial line to Inner Loop Controllers (ILC), response is read, added to response queue, added to command response queue and finally readout.
- Bitfiles - Contains the bit files that are loaded onto the FPGA's
- SettingFiles - Default configuration files
- src - Contains the C++ code for the ts_M1M3Support application
- tests - Contains python tests for the ts_M1M3Support application
- Utilities - Contains the code for the utility applications
This software depends on the fmt lib (MIT License), and users must comply to its license: https://github.com/fmtlib/fmt/blob/main/LICENSE.rst
Please use lsstts/mtm1m3_sim Docker container to run simulator.
BASE=$(pwd)/M1M3SS
mkdir -p $BASE
git clone [email protected]:lsst-ts/ts_m1m3support
git clone [email protected]:lsst-ts/ts_sal
git clone [email protected]:lsst-ts/ts_opensplice
git clone [email protected]:lsst-ts/ts_xml
source ts_sal/setup.env
export LD_LIBRARY_PATH=${LD_LIBRARY_PATH}:${BASE}/ts_sal/test/lib
cd ts_sal/test
cp -v ../../ts_xml/sal_interfaces/*.xml .
cp -v ../../ts_xml/sal_interfaces/MTM1M3/*.xml .
cp -v ../../ts_xml/sal_interfaces/MTMount/*.xml .
salgenerator MTM1M3 generate cpp
salgenerator MTM1M3 generate python
salgenerator MTMount generate cpp
salgenerator MTMount generate pythonThe base cRIO must be configured and up and running. OpenSpliceDDS shared libraries must reside in path searched for dynamic libraries. SAL libraries are statically linked into CSC binary, so those aren't needed on cRIO.
ExpansionFPGA resource name (set in NI MAX, under resource name) must match the name specified in ExpansionFPGApplicationSettings under resource. This URI also contains expansion cRIO IP address:
rio://139.229.178.185/M1M3-SupportExpansion-RIO
Please supply "DEBUG=1" to make call to compile with debug data (-g flag).
Please be aware that cRIO memory usually isn't sufficient to compile code with debug data. Compiling code without debug data, e.g. no DEBUG=1 option, works on cRIO.
Ni RT stack isn't needed.
cd $BASE
source ts_sal/setup.env
cd ts_m1m3support
makecd $BASE
source ts_sal/setup.env
cd ts_m1m3support
make ipkCopy and install ipk on cRIO with opkg install command (if needed, use --force-reinstall to overwrite previous package version).
If you run make before (for deployment), you need to run make clean before compiling for simulator.
cd $BASE
source ts_sal/setup.env
cd ts_m1m3support
make SIMULATOR=1Simulator does reasonable job computing various feedback signals. Those signals (values) and commended values are know to not be computed correctly:
-
HP measured forces those are hard to calculate. Direct consequences is that HP chasing doesn't work.
-
- HP chasing steps and mirror position as force computation doesn't work, HP and mirror position is not to be trusted in simulator as the mirror is raised or lowered. HP are send to target position after raise is completed (mirror weight is 100% completed), so mirror position is reported correctly when mirror is in active state.
With the exception of OpenSpliceDDS libraries, which needs to be installed from third party provided repository, opkg package contains everything needed to run the CSC on cRIO. Standard startup script is provided, so:
/etc/init.d/ts-M1M3supportd restartshall (re)start the CSC. In case you prefere to run (for debugging,..) CSC from the terminal, setup the environment by:
set +a
source /etc/default/ts-M1M3support
set -athen you can start CSC with:
ts-M1M3supportd -c /var/lib/ts-M1M3supportM1M3 command line client can be run (after you stop CSC) as:
/etc/init.d/ts-M1M3supportd stop # not needed if CSC isn't running
m1m3sscli
Please type help for more help.
M1M3 > helpSAL isn't needed to run the command line tool.
After make SIMULATOR=1, you can run the code as simulator. This doesn't need any hardware attached.
source ./ts_sal/setup.env && ./ts_m1m3support/ts_M1M3Support -c SettingFilesThe simulator does a pretty good job giving realistic data, you will notice some data isn't populated, it may be added at a later date, it may not. When issuing state change commands make sure to specify "value" as otherwise the command will be rejected.