I think the recent issues with squeezer pump ISS railing are likely resolved after today's on-table realignments. The green pump AOM was significantly clipping the beam (SQZT0 layout here, "GAOM1"), both reducing the total power after the AOM, and degrading the fiber coupling efficiency due to the clipped+bad mode shape. See GAOM1 sweeps from this morning (before aligning, left) and now (after, right). Along the pump path, we now have better GAOM1 throughput (~90%), comparable diffraction efficiency (~60%), and improved fiber coupling. For 60uW opo trans we've been using, we previously launched 22mW with no buffer (hence iss railed); now we only have to launch 15mW for the same opo transmitted power, with lots of buffers.
GAOM1 typically has ~90% throughput, e.g. previously this was 14.4mW out from 16mW incident. Today I first measured 25mW out for 44mW incident (~57%): this means >30% of power from SHG was clipping on the AOM apeture. After re-aligning GAOM1 using the mount screws, we're back up to 40mW out for 45mW incident, so almost 90% throughput again. I think this could be better, but good enough for now.
Relieving the clipping also significantly improved fiber coupling efficiency, presumably b/c it improves the mode shape, as we are using the 0-th order AOM mode for pump light, and not the nice diffracted order that gets cleaned up in the diffraction. So aom clipping shows up on the beam we are trying to fiber couple.
-- Before, for 20mW into the pump fiber, we had 1.8mW on OPO_REFL_DC_POWER on the other end, with GAOM1 driven at 2-3V (almost no room), and all power on that path going into the fiber (SHG_REJECTED = 0mW).
-- After, for 20mW into the pump fiber, we have 2.9mW on OPO_REFL_DC_POWER (60% higher power), with GAOM1 driven at 4.5V (mid-range), and a healthy buffer of rejected power on the fiber path (SHG_REJECTED = 8mW).
I've upped the generated sqz levels, back to more "normal", hopefully nominal, values. In $(userapps)/sqz/h1/guardian/sqzparams.py, Line 12, I've set opo_grTrans_setpoint_uW = 85, and re-tuned the OPO co-resonance temperature at this pump power (temp decreased from 31.7C to 31.68C, expected as the higher power heats the opo crystal more).
This OPO green trans setpoint of 85 uW corresponds to a generated SQZ level of about 15 dB, or non-linear gain of 12.3. By comparison, 60uW ~ Gen SQZ = 12dB ~ NLG 6.4. So, this is nearly doubling the squeezer gain, aka increasing the generated squeeze level by about 3dB.
To check whether pump ISS failures are from aom clipping in the future, we can try the following. SQZT0 medm stuff should look like this screenshot, with teal circle highlighting the path we're talking about, and red arrows showing some knobs to turn.
1) Bring SQZ_OPO_LR guardian "DOWN". Now PUMP_ISS is disabled, so GAOM1 box is yellow.
3) Turn off pump fiber flipper.
2) Set H1:SQZ-OPO_ISS_DRIVEPOINT = 0. This turns off GAOM1 diffraction, and sends all power to the fiber (pump light uses 0-order).
3) See the total pump power after GAOM1 (sum of H1:SQZ-SHG_REJECTED_DC_POWERMON + H1:SQZ-SHG_LAUNCH_DC_POWERMON).
4) If GAOM1 is well-aligned and not clipping badly, I'd expect total power after GAOM1 to be about 70% of the SHG output green power, H1:SQZ-SHG_GR_DC_POWERMON.
Broken down, this ratio is from SHG's green output light (H1:SQZ-SHG_GR_DC_POWERMON) being split ~25% to FC green locking, and the remaining ~75% to the pump path GAOM1 which lets through ~90%.
I'm a bit suspious how this happened in this first place: I wonder if the beam is slightly large for the AOM aperture (but aom's throughput is not terrible), whether SHG or lab temperature drifts are changing the output alignment or mode-shape slightly, or whether we've bumped the SHG output path doing other on-table work. If ISS railing becomes a problem again at this SHG power, we can look into it more. For reference, we last increased SHG power by ~15% on May 16th (69671) in an attempt to resolve exactly this issue; we don't need all this shg power anymore.
Attaching quick SHG, OPO, CLF transfer functions taken after table alignments, with pd powers on medm.
Just following up on trends since this on-table AOM alignment -- this on-table fix seems to have mostly resolved the previous squeezer pump ISS issues.
