We had Sheila's ring heater change on for ~4 hours, and it seems to have prevented the PI from coming up, which is great.

I have now changed all 4 test mass ring heaters to be 1.7 W per segment, starting at 23:45 UTC.  I didn't use the filtering guardian, I just punched in the numbers.

Why 1.7 W per segment?  It's a rough guess:

• Peter pointed me to a 2009 document (T0900359) where they were doing some modelling to determine whether we should coat the barrels of our ITM compensation plates (spoiler alert: we didn't do it).  It has some useful modelling for what happens to both the front surface and to the optical path when travelling through the ITM substrate.  So, each of the sets of plots shows roughly what will happen to the arm cavity mode, and what will happen to modes in the vertex. 
• All of the plots are meant to show how very nicely we can do if we use both ring heaters and annular CO2 heating.  However the table on page 3 shows that you can still improve your situation if you use only ring heaters and no CO2.  The metric is the 'correction factor', which is some kind of ratio to indicate how much light didn't get scattered out of the TEM00 mode if you do use TCS, versus if you don't use TCS.  Higher is better, with a CF = 1 being exactly the no TCS situation. That table shows that you can still get a correction factor of 10 if you use only the ring heater. 
• The plots on page 3 and the table on page 4's specific numerical values for the settings are for the case of having a coating on the barrel of the CP.  The plots on page 5 are for the case that we actually have, of no extra coatings.
  • The table on page 4 indicates that if only using ring heater and no CO2 (last row), you need 2.7x ring heater versus the case where you have both ring heater and CO2 (first row) (10.5 W / 3.9 W).
  • For the case that we have, taking the ring heater value from the plots on page 5, I want to put 2.7 * 6.3 W on each test mass.
  • However, all of the values assume that you have 500mW absorbed on your optic, but we have more like 100mW with our current IFO powers, so we need to divide by 5:  (2.7 / 5)* 6.3W  (2.7 / 5)* 3.9W
  • But, that's a total power per optic, which I need to spread over the two halfs of the ring heater:  (2.7 / 5 / 2) * 6.3W = 1.7W per segment  (2.7 / 5 / 2) * 3.9W = 1W per segment

So far, in the first 15 minues, most of our buildups are decaying.  But, I plan to leave these values in place, and we'll see what happens over a longer term.

I am leaving the IFO requested to DOWN, so that if it does lose lock, it won't try to reacquire on its own.

EDITED to clarify that I was wrong - we do have gold coatings on the compensation plates.  Alog 57662, for example, shows our new ITMY with its gold coated CP.  I had used 1.7W per segment, but I should have used something more like 1.0 W per segment.  However, I'm not 100% sure that that value will be correct for the ETMs, since this modelling was just for the ITMs.  Perhaps we'll give this a try, since using the 1.7 W values caused buildups to go down, and we lost lock not too long after having turned on all the ring heaters.