Sheila, with many conversations with Keita, Jeff K, and others
We have several reasons to believe that something changed in our suspensions durring the Montana EQ. (See alog from Beverly Berger and Josh Smith, (37775) which got us started looking at this, and Cheyl's log about the large triples, (37674). We are still looking at some of the data from alignment sensors, but here arer some things we can say:
- The torque required to produce a fixed top mass pitch has changed after the EQ.
- The 1st attachment shows the top mass applied force (calibrated in urad using sliders) vs osem readback for all 4 quads, color coded for before and after the EQ. The linear relationship before the earthquake is offset from the relationship after the EQ. The torque applied to the top mass are calibrated into urad using the slider calibrations, I will update this alog in the morning with torques in torque units.
- The top mass pitch required to produce the same test mass alignment is 100s of uradian different before/after the EQ.
- We trust the oplevs because restoring them after the EQ allowed us to lock the IFO with roughly the same spot positions as before the EQ. (37451). The second attachment shows you the change in top mass pitch required to produce a similar opLev pitch for each test mass.
- We think that the osems (at least on the quad top masses) are OK.
- Jeff Kissel's top to top transfer functions which look the same as before the EQ. (37689)
- The top mass actuation strength seems to be the same as before the EQ. In the first attachment the slope of the linear relationship before the EQ is the same as the slope after, meaning the actuation strenth is the same.
There is more to be done checked on here, for example, checking if this has happened at any other times during the run (I checked one large EQ, I see no shifts like this), checking yaw (which has much smaller shifts than pitch), checking the triples, and looking at the alignments of the reaction chains. Can we interpret the information we have to make a gues aout what might have changed in the suspension? Wire slipping or some kind of damage to the prisms are some things we have been thinking about.
Hysteresis is a possibility here. We discovered on the LASTI quad during one of the early builds that these suspensions have significant hysteresis in pitch. That is, if you tip the stages a given amount, they will not come back all the way, leaving you with a pitch offset. The attached plot shows a measurement of this effect from LASTI, showing a pretty standard looking hysteresis plot. We learned that you can 'undo' any offsets by getting the pendulum swinging, and letting it slowly damp itself. The slower the ringdown, the closer it returns to its nominal 'equilibrium' position.
The offsets you see here don't look any bigger than what we saw there, though granted we were trying to measure the effect, so pushed it pretty far. Then again, we didn't have any major earthquakes either.
There were many documents written to investigate what we saw at LASTI. Mark Barton's document, T0900103, includes a list of most if not all of them.
So it could be that this earthquake induced some hysteresis offset, or perhaps there was an offset already and the swinging motion from the earthquake removed it. Anyway, try swinging the pendulum in pitch with some large *but safe* amplitude, and you should return to the nominal 'equilibrium' position, if it isn't already there.
Looking at Beverly log (37775) that shows DC changes in the pitch offset across the earthquake time. Are the changes in the pitch in the lower masses compatible with the reported change at the top mass?
Here are some additional plots, for those who are interested in what happens to the osems between the reaction mass and the top mass. I also have plots that show torque applied to the reaction mass vs measured pitch, these aren't very useful because we don't change the torque applied to the reaction mass, but they do show that there were similar shifts in the reaction chain. In order to interpret the data from the L1 and L2 osems we will need to account for shifts in the reaction chain.
Posting a jpeg version of the LASTI hysteresis plot above, since the pdf was causing issues.
Also, here is a summary of the procedure I used to make the plot way back in 2008:
"These data points are separate pushes and releases. The procedure was to put the top mass on its stops with the rest of the chain suspended. Note, the quad was on the test stand outside the chamber at the time. Then the top mass stops were used to tip the top mass some amount in pitch. The angle of the test mass, with the top mass still tipped, was measured with either an autocollimator or optical lever. That test mass angle is the X axis in the plot, called ‘Input Pitch’. Then the top mass was released slowly to avoid oscillation, and the test mass pitch angle was recorded again. That value is the Y axis, called ‘Output pitch’. This process was repeated for successively larger and larger Input pitch values, until I was afraid to tip the suspension any more. I then started to tip the suspension in the other direction until I was again afraid to tip the suspension any more. And finally, to close the hysteresis loop, I repeated some of the data points along the original tipping direction."
