J. Kissel, B. Weaver

One of the actions left on the HAM5 to-do list (see recent snapshot in LHO:60434) is 
    "0%    237    Swap BOSEMs - refurbish"
which is in reference to the long-standing IIET ticket 10093: BOSEM LED decay. For context -- because it was a concern during the design phase, we got worried after trending many suspension's OSEM sensor values over the course of the past many years of aLIGO (see e.g. LHO:55221) and seeing a slow downward trend.

However, since then, after a big long review of the topic T2100112, the conclusions remain:
    - just trending the OSEM displacement values is not a good enough metric to really find the smoking gun of LED decay, since it's confused by many other things including alignment (both physical and requested) and temperature,
    - the trends send in that data are visible via a zoomed in range that is not the same from OSEM to OSEM, so trend *rate* can be misleading just on the zoom. A fair-comparison, computing the *relative* change across the OSEMs show that -- for those OSEMs that *do* show a downward trend -- the drop is at most 5%
    - A noise budget of the OSEM electronics show that the system is already pretty low in terms of signal to noise ratio, so a 5% drop in current signal *and* noise won't significantly impact the performance
    - It would be a ton of work to whole-sale find other, satisfying quantitative metrics that would definitively say "we should swap this and that OSEM because it's LEDs have decayed to too low," and we just don't have the person-power to do that work
    - And thus -- (though mostly from "stalemate") -- the conclusion from that review is that no immediate action need be taken that is different from "normal." Where "normal" is to ask a human to assess if a BOSEM (or AOSEM) needs replacement when a vent opportunity arises.

Today, that opportunity has arisen for SR3 and SRM in HAM5.
In this aLOG, I cover my recommendation for SRM.
In the comments, I'l cover my recommendation for SR3.

Here's my human assessment of H1 SUS SRM, based on the following (very limited) data (copied and expanded upon from the "LHO tab" of Stuart and Betsy's assessment E2000472): 
    OSEM       Jul '18    Feb '20    Sep '20   May '21     OLC           OLC            % Diff
               During     During     During    During      Test          Final          (test - final)
               NLN        NLN        NLN       NLN         Electronics   Electronics     / test
                                                           Circa '12     Circa '14
                                                                         LHO aLOG 60477
    SRM M1 T1  16500      16200      15200     14528       27900         26908	         3.56       
    SRM M1 T2  16000      15800      15900     15550       26200         25472          2.78		
    SRM M1 T3  13000      12900      11500     10868       29600         28670          3.14		
    SRM M1 LF  18718      17858      18200     18179       28850         27764          3.76		
    SRM M1 RT  8200       7900       6200      6389        27000         26272          2.70		
    SRM M1 SD  14000      13300      12940     12790       26750         25516          4.61
From the above data, we really can't conclude much.
    -- Remember, only T2 & T3, and LF & RT are involved in Pitch and Yaw, respectively, alignment of the optic.
    -- T2 and RT have changed quite a bit, dropping ~20% from 2018, but their pairs T2 and LF have not dropped barely a few percent. This differential drop implies that pitch and yaw have been requested to change, or less likely, out of some extreme coincidence, these two OSEMs (T2 and RT) have dropped
    -- Also not that the large change in RT (i.e. yaw OSEM) only occurs between the Feb '20 and Sep '20 period, which is the exactly the ITM swap, and we *definitely* had to find a whole different DC alignment of all the optics because of that. 
    -- Since the Sep '20 change, the May '21 progression actually went *up* which is also consistent with an alignment request change rather than an OSEM decay.
    -- In terms of the change in open light current between 2012 and 2014, (a) there's very little change, and (b) that change is perfectly consistent with the changes we might expect from switching from test stand electronics to production electronics. We'd need a  new measure of the open light current with the same electronics to really conclude anything truly quantitative. 

Also through the course of the past several days of research, I've concluded that -- over the 0.7 [mm] viable sensing range of the OSEM, the actuation strength remains relatively constant. As such, we should be able to get the same alignment position from the same DAC drive requested (and thus alignment slider request) regardless of whether we change the centering of the OSEM.

As such, here's what I recommend for SRM: in short -- just recenter the entire suspensions OSEMs, and do not replace any BOSEMs.
Do so with the following suggested careful process.
    - Turn on damping loops and alignment sliders to reproduce a good alignment (according to M2 and M3 OSEMs) from a time during May '21's nominal low noise time.
    - Record values for M1, M2, and M3 OSEM positions (in terms of OSEMINF INMON ADC counts *and* OSEMINF OUTMON microns).
    - Record DAC output request for the given alignment slider value.
    - Recenter M2 UL, UR, and LR AOSEMs *without remeasuring the open light current, nor updating OSEMINF gains or offsets*
    - Recenter M3 UL and LR AOSEMs *without remeasuring the open light current, nor updating gains or offsets*
    - Record newly centered values of M2 and M3 OSEM positions with damping loops and alignment sliders ON.
    - Turn OFF damping loops and alignment slider values.
    - Record not-yet-centered M1, and newly centered M2, and M3 OSEM positions with alignment sliders OFF (confirming they're still in range).
    - Back off *all* M1 OSEMs, and remeasure open light current values, but *do not update the OSEMINF gains or offsets*.
    - Recenter all OSEMs, but **account for the change in position, between alignment sliders *on* vs *off*.**
    - Turn on alignment offsets, and confirm that you've done math right, and the M1 OSEMs are indeed centered, and confirm that the M2 and M3 OSEMs are in the same position.

In this way, we'll have ensured that the OSEMs can recreate the alignment that we need during nominal lownoise, and the sensor side of the OSEM will be nicely centered around that value, and we'll have not changed anything else mechanically about the suspension, so it should have the same range surrounding that alignment as before.