The temporary type B guardian now has the following states. See typebgrdtemp.png for the state graph and the edge transitions.
| State | Type | Description...................................................................................... | Behavior...................................................................................................................................................... |
| INIT | goto | Initial state | Does nothing. |
| TRIPPED | goto, redirect = false | Goto this state if watchdog is tripped | Loop while watchdog is not reset. |
| RESET | goto | A goto state to pass through whenever SAFE state is requested. | 1) Set all output ramps to 10. 2) Set all output gains to 0. |
| SAFE | request | Disable all output. | 1) Set all output ramps to 10. 2) Set all output gains to 0, if they were not off already. 3) Turn off master switch |
| MASTERSWITCH_ON | transition? | Turn master switch on. | Turn master switch on. |
| TOWER_DAMPING | transition? | Soft-start IP, GAS damping filters (and DC control) with 20 seconds ramp and then change the ramp to 5 seconds for better control performance. | 1) Disable input, output and offset and then reset the controller buffers. 2) Restore offsets that are stored in $(ifo)vis$(optic)_align.snap (now the take align snapshot action also stores offset in DAMP filters) 3) One at a time, set IP_DAMP_{L, T, Y}, {F0, F1, BF}_DAMP output ramp to 20. 4) One at a time, set IP_DAMP_{L, T, Y}, {F0, F1, BF}_DAMP output gain to 1. 5) Wait for 10 seconds 6) Set all ramp to 5 seconds |
| TOWER_DAMPED | request | With all IP and GAS degrees of freedom set to a known good position for alignment, again, stored in $(ifo)vis$(optic)_align.snap. | Now it does nothing. But I will probably add a line to make it continuously monitor the watchdog. Another possible improvement for the future is to check for saturated output which could set the control system into an unstable state and jump to a previous state and notify the user for further actions to be taken. |
| PAYLOAD_DAMPING | transition | Engage IM damping filters with 5 seconds output ramp. | 1) Set all IM DAMP output ramp to 5. 2) Set all IM DAMP output gains to 1. |
| PAYLOAD_DAMPED | request | With IM damping filters on. | Same as TOWER_DAMPED. |
| PAYLOAD_RAMPING_DOWN | transition | Turn off IM damping filters with 10 seconds ramp | 1) Set all IM DAMP output ramp to 10. 2) Set all IM DAMP output gains to 0. |
| ALIGNING | transition | Roughly align the optic with OPTIC ALIGN offsets (P and Y) stored in $(ifo)vis$(optic)_opticAlign.snap. | 1) Restore offsets (P and Y) that are stored in $(ifo)vis$(optic)_opticAlign.snap. 2) Enable OPTIC ALIGN offset, and output if they were turned off. 3) Set OPTIC ALIGN ramp to 20 seconds. 4) Set OPTIC ALIGN output gain to 1 (P and Y simultaneously). |
| ALIGNED | request | With the optic aligned with OPTIC ALIGN offsets. | Same as TOWER_DAMPED |
| MISALIGNING | transition | Same as ALIGNING but with offsets stored in $(ifo)vis$(optic)_opticMislign.snap. | Same as ALIGNING but with offsets stored in $(ifo)vis$(optic)_opticMislign.snap. |
| MISALIGNED | request | With the optic aligned to a known safe misalign direction. | Same as TOWER_DAMPED |
| DISABLE_ALIGN | transition | Disable OPTIC ALIGN with 10 seconds ramp. | 1) Set all OPTIC ALIGN output ramp to 10. 2) Set all OPTIC ALIGN output gains to 0. |
See state graph here or attachment below.
The Guardian system has been tested roughly with the SRM suspension.
One thing to note is that the guardian does not handle the dc control very well. For example, the actuator could easily be saturated if the set point (negative offset) is too far away from the suspension's position and Guardian would do nothing about that. I could implement safety features to prevent that from happening. But still, the guardian can never get past that point where the actuator gets saturated, and eventually ends in an endless loop. So, please take align snapshot with IP and GAS set points close to the current position of the suspension. Note that this problem is also temporary only. This problem will not be a big deal after the new model changes because the guardian could easily ramp the set point from the current position to the designed position.
This version is a temporary one because it is only compatible with the real-time model that is currently in use. In the future, new blocks/filters will be introduced and the flow of the control system will be changed. So, the Guardian code (along with Burt request files for the align states) will be changed accordingly in the future. In the new configuration, we would be able to 1) run dc control and damping control in parallel (as opposed to running in series, which is wrong, and is still the case in ALL suspensions control systems.), 2) disable the "-1" filter which is implemented in every control blocks, 3) have a REAL set point (as opposed to offset) with adjustable INPUT ramp (we have been using output ramp, which is not a good strategy for control system.). The new real-time model for the type B tower is almost ready. The real-time model for the payload was not changed (yet) because it is still unclear whether a DC control (IM to TM) will be needed. However, new changes have not been applied yet since the suspensions are in the restoration process and will be used in commissioning soon. The changes will be made when there is a window where the suspensions are not in use and DAQ reset is available.