We reached 50W relatively stable and reliable, and were able to use the POP A offsets to drag the interferometer around.

I recorded that particular alignment vector, and stuck it into a script. This now allows moving the full interferometer in this degree of freedom very rapidly, without taxing the soft loops too much.

We see a clear recycling gain dependence on this degree of freedom. However, one oddity is that when we reach the maximum recycling gain, the AS90 buildup becomes very ratty. I am wondering whether we get close to clipping somewhere with this move, or whether it is related to our SRM alignment dither.

The alignment vector for this move is

# pit
gppr3=0.07
gpprm=-13.2
gpim4=-325
gpsr2=-2.39
gpsrm=3.2
gpex=-0.7
gpey=-0.7
gpix=-0.45
gpiy=-0.35

# yaw
gypr3=0.028
gyprm=11.07
gyim4=61.22
gysr2=2.22
gysrm=2.74
gyex=-0.5
gyey=+0.5
gyix=+0.32
gyiy=-0.31
 

Sheila, Stefan

Based on yesterday's measurement, the ASC PUM to TOP x-over was at 0.002Hz and 0.0029Hz (before the 10dB increase in the script).

I added gains of 50 for pitch and 34.5 (for now in the L1_LOCK modules) to move this x-over to 0.1Hz.

The only loops for which this 0.1Hz offload is critical are the soft loops. Thuis I added z=0.01Hz:p=0.1Hz lead filters to all soft loops. I also turned the soft loop bandwidth up by 3x.

The guardian now sets those gains before turning on the soft loops - and sets them back in the down state.

We still should do some more loop tuning for the soft loops - their response should be sped up

We got to 50W - but lost it when moving the PRM - the soft loops seem to be still too slow to follow, resulting in the Izumi-Ballmer instablility.

Stefan, Sheila

We had a lockloss last night during the INCREASE_POWER state when the ISS third loop was coming on.  This is done by a function that had a ten second sleep in a while loop, so the guardian didn't recognize the lockloss for about 90 seconds, in which time ETMY and ITMX suspesnions tripped.  Stage 2 of both ISIs tripped after the suspensions had been tripped for about an hour.  Screenshot shows the sequence of events, first a 90 minute time span then in a 2 minute time span around the lockloss. 

To make sure that in the future the guardian will recognize the guardian more promptly, we moved the ISS third loop engagement to the IMC_LOCK guardian which currently handles the ISS second loop engagement.  The ISC_LOCK guardian just requests this now in the increase power state when it reaches 10 Watts.  

We also looked at why the lock broke.  One potential explanation is that there are large glitches in both DHARD P and Y when the boosts get engaged about 3 seoncds before lockloss (screenshot attached).  These are simple single pole single zero boosts, which were Always on with a 3 second ramp for engaging.  We changed the ramps to 0.2 seconds to see if by ramping faster than any features in the filter we can avoid glitches.  The real cause of the lockloss might have been changes to the offloading that weren't compatible with the guardian as it was last night, but in any case we would like to avoid these large glitches when engaging the boosts.  

I started to look into a faster ASC UIM to TOP stage offloading.

The DC gain ratio of the DRIVEALIGN filter outputs is about PUM:TOP = 23:20000 (pitch) and 30:20000 (yaw). There is a slight mismatch between the optics, but that seems to be taken care of with the CAL filter (FM1 in M0_LOCK).

With that gain ratio, and with FM1, FM4 and FM5 on in the MO_LOCK banks, the current x-over between TOP and PUM is about 0.002Hz (pitch) and 0.0029Hz (yaw).

This is very low. I verified that with 10x the gain in all 4 soft loops (gain=0.45 instead of 0.045) and 10x the offloading gain to top, the soop[s still seem stable. I believe the reason we are not running with that during the angaging these loops tax the SRC loops.

I added 0Hz:0.1Hz AC couplers into the L2 DRIVEALIGN P2P and Y2Y banks, as well as 0.002Hz:0.1Hz lead filters into Mo DRIVEALIGN P2P  banks ( 0.0029Hz:0.1Hz lead filters for Y2Y). These filters were successfully engaged in full lock. Somewhere however the guardian turns up the offloading gain by 10dB, making these filters unstable... So for now they are not in guardian.

Unfortunately, the pwer-up is still failing randomly. Not sure whether that's the 40mph wind, the numerous locked-down hepi's or the increased noise on RF DARM.

Temps look good -> made minor variac adjustments to midstation-destined RGA components that are currently "grafted" onto VBOB (exploiting available clean pump system to bake this sub-assembly and to get a preliminary look at the suspected "dirty" Nitrogen and Krypton calibration gases following bake cycle)

Jenne, Stefan

- Adjusted the PRC1 input matrix form REFL to match the input matrix from POP X at 2Watt. THuardian now has a switch : lscparams.use_popx_wfs (1= use POP X, 0= use REFL WFS). Left it on POP X for now.

- Adjusted the dither input matrix for the SRM to maximize the ratio AS90/POP90. This was done with the gains -0.013 for AS90 and 0.5 for POP 90. (The gain ratio is simply the DC ratio of POP90 and AS90: 17:650.)

- We next tried to establich a stable locking point for the PRM alignment - but unfortunately repeatedly lost in during power-up.

I've had a nagging in the back of my mind for a little bit now, so I looked at the input steering mirror positions from a good time during O1 as compared to now.  This doesn't account for everything yet, since I haven't looked at the pointing of the IMC, but the current OSEM readbacks of the IMs are significantly different than they were during O1.  I wonder if part of our whole problem has been that we just have different enough input pointing that we're having PRC gain issues.

Anyhow, I'll meditate more and look at IMC pointing, but then maybe I'll try moving the IMs back to these Sept 24-29 positions on Sunday.

As a proxy for PRCgain since we didn't have the handy channel that we do now, I plot the TR_X_NORM.  On the left side of the screenshot are the September values, including the TR_X_NORM at about 1320 counts.  When we are at 50W and our (new) PRC gain channel reads 25, TR_X_NORM is at about 900 counts.

Carlos, Dave, Stefan:

earlier this afternoon digital video cam04 (ALS X) and cam18 (AS AIR) developed problems. Through a combination of power cycling the cameras (via POE on the network switch) and restarting the python programs these were recovered. This evening Stefan found cam22 (ITMX Green) also had an issue. Its image was changing, but its corresponding epics channels were not updating and the date/time on the image was frozen. We tried restarting the CAM22 process on h1digivideo2 (manually and via monit-web) but the issue did not clear. I then rebooted h1digivideo2 (it had been running since 06Jun2016) and the problem cleared. We saw CAM04 had developed the problem again, so I rebooted h1digivideo0 which cleared this. We decided not to reboot h1digivideo1 as its cameras are working correctly.

After I and Evan relocated POP-X and POP-L2, Jenne increased the ASC gain of POP-X by a factor of 250 for PIT and 500 for YAW (alog 28666) just to get back to the old UGF, which sounded crazy to me as I didn't expect much change in Gouy phase.

Just to see if my assumption was wrong and we got super unlucky, I calculated the Gouy phase of the POPAIR path, and it seems like there shouldn't be much change.

In the first attachment, left column is the current configuration, right is the old one. Bottom is the entire POP path from ITM to the ISCT1, and top is the zoomed-in view from HAM1 to ISCT1.

POP-L2 was placed far enough from the waist originally. This lens was moved farther from POP-L1 by 4 inches later, which should have increased the Gouy phase, but it's only 9 degrees. The beam diameter is 2mm now instead of 4mm but that should be OK. These don't explain the crazy decrease of the optical gain.

Funny thing is that the POP-X spectrum itself looks almost the same before and after the relocation (second attachment), so I'm kind of dubious that the optical gain is lost.

It's not totally impossible, but very unlikely, that the Gouy phase was not great to start with, e.g. 81 degrees for the DOF we want to see, and after the change it became 90 degrees.

Noteworthy only in that this unit had been dormant for the past year or two -> Pumps running and hot surfaces

Over-filled CP3 with the exhaust bypass valve fully open and the LLCV bypass valve 1/2 turn open.

Flow was noted after 41 seconds, closed LLCV valve, and 3 minutes later the exhaust bypass valve was closed.

In looking at frame writer stability issues I installed a daqd build on h1fw2 that would report some timing information on the main producer thread.  This is the thread that reads the data in the data concentrator and checksums it and queues it up for the frame writing threads.

The four points I am measuring are:
full - a full cycle (should be under 1/16s)
recv - time to receive a 1/16s block of data
crc - time to checksum the data
xfer - time to move the data into the frame writing buffers

These numbers are do not include the time taken to output the timing samples to the log (so the full cycle time was really a bit longer).

Here is a particularly bad sample, time is in seconds:
 full min:0.030165 mean:0.0694159 max:0.371219
 recv min:0.00241995 mean:0.0398388 max:0.338558
 crc  min:0.0151799 mean:0.0171485 max:0.020812
 xfer min:0.0118489 mean:0.0124278 max:0.0145991

And another that was more typical

 full min:0.0603912 mean:0.0624495 max:0.0659258
 recv min:0.0284889 mean:0.0316754 max:0.0346711
 crc  min:0.0155029 mean:0.0183311 max:0.021708
 xfer min:0.012187 mean:0.012442 max:0.0149119

When a cycle on the producer thread goes over 1/16s we get retransmission requests on the network.  I had to stop running this build of daqd due to it triggering a large number of retransmission requests.

During a recent meeting at LLO to discuss daqd Keith had informed us that the producer thread is doing to much and it needs to be split into separate threads.  The timing test I ran today supports this.  I am working on code now to split the producer thread into two halves.

* reminder h1fw2 only writes to local disk and is being used to test fw changes under a production data load.

Evan G., Jeff K., Evan H.

Summary:
We have updated the L2-L3 crossover filters to reduce the effect of the L2 actuation on the superactuator at frequencies above the L2-L3 crossover. In addition we have turned off the vStopB filter in the L2 drivealign bank. The old filters remain in the filter bank, but are turned off in case we need to revert. The SDF has been updated to reflect these changes.

Details:
The old lowpass [ :30], and highpass [0:30] filters were allowing too much interaction of the L2 stage on the frequencies above the crossover. We needed more aggresive rolloff of the L2 stage. A new design was used to develop prototype lowpass and highpass filters that were then normalized to produce complementary filters. The trick was to add some additional filtering in addition to the single pole or single-zero/pole design.

The Matlab zpk prototype lowpass is:

    29.804 (s^2 + 792s + 4.077e06)
  ----------------------------------
  (s+94.25) (s^2 + 1192s + 1.304e06)

and the prototype highpass is:

  0.68377 s (s+0.4816) (s+40.81)
  -------------------------------
  (s+125.7) (s^2 + 35.75s + 1174)

The resulting complementary filters are the lowpass:

      43.587 (s+125.7) (s^2 + 35.75s + 1174) (s^2 + 792s + 4.077e06)
  ----------------------------------------------------------------------
  (s^2 + 27.96s + 1124) (s^2 + 256s + 1.986e04) (s^2 + 1087s + 1.175e06)

and the complementary highpass:

        s (s+94.25) (s+40.81) (s+0.4816) (s^2 + 1192s + 1.304e06)
  ----------------------------------------------------------------------
  (s^2 + 27.96s + 1124) (s^2 + 256s + 1.986e04) (s^2 + 1087s + 1.175e06)

Putting this into Foton yields the lowpass design:

zpk([63.027+i*315.14;63.027-i*315.14;2.8446+i*4.6522;2.8446-i*4.6522;20], [20.375+i*9.3751;20.375-i*9.3751;2.2252+i*4.8498;2.2252-i*4.8498;86.475+i*149.24; 86.475-i*149.24],1,"n")

and the highpass design (properly normalized to 1 at high frequency):

zpk([94.821+i*155.07;94.821-i*155.07;0;0.076645;6.4948;15], [20.375+i*9.3751;20.375-i*9.3751;2.2252+i*4.8498;2.2252-i*4.8498;86.475+i*149.24; 86.475-i*149.24],1,"n")gain(0.000578996)

These are installed in the L2 and L3 drivealign banks as L2L3LP and L2L3HP, respectively.

We have compared the old and new design and the effect on the DARM actuation and DARM open loop gain and closed loop suppression, and we are satisfied with the improved filter design. Attached is a series of figures to illustrate the new design (see the plots.pdf file)

Figure 1: actuation plant transfer function (left plots) and the PUM/TST ratio (right plots)
Figure 2: current state of the distribution filters (before any modification)
Figure 3: current state of the actuation authority (before any modification) -- note the black curve has ripples in the 100-900 Hz band
Figure 4: new crossover filter design -- prototype and final design
Figure 5: new distribution filter design from Matlab continuous filters -- note that we turn off the vStopB filter module in the L2 drivealign bank
Figure 6: new actuation authority from the Matlab filters -- note the improvement in the black curve with much less ripple than before
Figure 7: using new design, L2 and L3 actuation transfer functions: drive at input to L3 lock bank and look at L3 response for both L2 and L3 paths and taking the sum
Figure 8: using new design, the ratio of the two single curves in figure 7 -- note the transfer function is unity at 23.3 Hz and with a phase margin of 67 degrees
Figure 9: new distribution filter design as implemented in Foton -- note that these are the same as the Matlab design
Figure 10: new actuation authority from the Foton filters -- again, same as before
Figure 11: compare the DARM open loop gain and closed loop suppression from the current state (old) and using the new filter design and also when there is no SRC detuning -- note there was a lingering 400 Hz notch filter in the ER9 run that shouldn't have been there (Evan H to make sure this is off) and this model takes into account the increase of the digital DARM gain to 1400. The blue curves have little ripple above 100 Hz, and the phase margin is actually a little larger than before, about 48 degrees. The gain peaking is roughly the same as before ~4 dB.
Figure 12: A final acutation authority plot showing the new total is much improved over the old total, especially in the range of 100-900 Hz (see zoom on right hand plots).

For the record, also attached are screen shots of the L2/L3 drivealign filter banks before and after the changes as well as a screenshot of the SDF.

State of H1: earthquake, unlocked, ISC_LOCK set to Down, IMC_LOCK set to Offline

Activities:

• HAM6 vent prep
• EY ion pump power supply issues - update: power supply is off for the weekend
• EY leak testing

Since EQ:

• 21:50UTC - Gerardo to EX - and back at 22:35UTC
• 22:00UTC - Marc in LVEA - at HAM5/6 terminating cables - out as of 4:10PM PT
• 22:40UTC - Jenne/David in LVEA - looking for Marc - out as of 4:10PM PT
• 22:42UTC - RichA/guest in LVEA - prep for filming in the LVEA tomorrow - still in LVEA

Following the same idea as elog 28442, here is a script for moving the beam spot position on PRM (prmspotmove.py).

It monitors IM3 slides, and moves IM4, PRM and PR2 sliders in such a way the the beam spot positions on PR2 and PR3 don't change.

The script was tuned in INPUT_ALIGN (for IM4 and PR2 matrix elements) and PRM_ALIGN (for PRM matrix elements).

Matrix  elements (in rad /rad IM3)

pitIM3toPR2=+0.0022;
yawIM3toPR2=+0.0055;
pitIM3toIM4=-1.3;
yawIM3toIM4=+0.92;
pitIM3toPRM=-0.009;
yawIM3toPRM=+0.031;
 

Also attached is the python script used for finding the matrix element, getMx.py.

Kiwamu, Stefan

Using the pr2spotmove script (see alog 28420), we moved the spot on PR2 in lock by almost 1 milimeter.

Conclusions:

- The carrier recycling gain desn't care about the PR2 spot positon.
- The POP beam from PR2 to the invacuum POP diode is fairly close to the edge of the visibility aperture. This might be an issue for LSC aux noise.

Details:

- To be able to explore the full range of motion, we had to switch back to the REFL B WFS for PRC2 / PR3 control. We BTW verified that this WFS also works at 40W (and during the power increase).
- With that, the move-limiting aperture is the path from PR2 to POP. We can move until the POP_A loses the beam. The pitch and yaw min/nominal/max values for PR3 alignment angles are
    Pitch value / delta:
     min: -302.95  / -4.6urad
     nom: -298.35 / 0urad
     max: -275.35 /  +23urad
    Yaw value / delta:
     min: 257.45 / -16urad
     nom: 273.45 / 0urad
     max: 281.45 / +8urad
  Notice how close to the edge we are in pitch - this could be a factor for PRCL/SRCL/MICH auxiliary length noise and scatter coupling. We have to revisit this in low-noise.

- The 27.6urad move in pitch (24urad in yaw) peak-to-peak move corresponds to 0.9mm (0.8 mm) for the PR2 spot position. Note though that for the beam in transmission of PR2 that corresponds to several spot sizes of motion. (The virtual beam waist behind PR2 is 114u, i.e. we moved the spot more than 7 beam waists. As a result the beam completly left the POPAIR camera view.)