I am so sorry that the last figure and text file in the above klog was wrong because I misunderstood the behavior of Guardian. 

This time, I define "lockloss" means being "DOWN" state without requesting it.

• a plot of where the lockloss occurred (Fig.1)
  • x-axis is a state of Guardian, y-axis is the number of events
  • There are many lockloss at "ALS_XY_LOCKED" state.
• a list of gpstime (lockloss_list.txt)
  • left is a state of Guardian, right is gpstime
• a list of how long it took in the state until lockloss occurred (duration.txt)
  • left is a state of Guardian, right is duration [sec]
  • It stayed 0.1875 sec at "ALS_XY_LOCKED" state until lockloss.

memo:

In the previous klog, I used "LOCK_LOSS" state to define lockloss. However, there are a lot of lockloss which don't pass "LOCK_LOSS" state. So I use "DOWN" state to define lockloss this time. Additionally, I took into account whether the lockloss happened intentionally or not, by using "K1:GRD-LSC_LOCK_REQUEST_N" channel.