Rana, Hang

Yesterday we saw the 0.47 Hz oscillation showing up in basically all the PIT ASC loops at 23 W and became unstable at 25 W.

We thus measured DSOFT P OLTF and the result was attached to this entry. The cyan/pink traces were some old measurements done back in October at 2 W. The OLTF measured today at 20 W was the blue/red traces (We have good measurement in the 0.3-0.6 Hz region yet below 0.1 Hz we could not get good). In both measurements  we saw a pair of zeros at 0.47 Hz, which was the freq we saw the oscillations.

To compensate for this feature, we put resgains at 0.47 Hz with Q of 3 and height of 12 dB in DSOFT P (FM9), DHARD P (added to the boost10W FM5), and CHARD P (added to boost10W FM5). The same filter was also created for CSOFT P yet it didn't seem necessary to engage it for now. After these changes the 0.47 Hz feature was basically removed from the PIT loops at 23 W.

An oscillation at similar frequency also appeared in CHARD YAW, and it was mitigated if we turned off the FM4 boost10W filter in CHARD YAW.

With those modifications we did not see oscillations at 23 W any more.

====================================================

In addition, we also improved our DSOPT/DHARD PIT decoupling for the B path (AC coupled; for damping the dPdTheta)

The new matrix we used for DSFOT P B path input was

In the second attached plot we compared the performance of this new input matrix to the old one (routed to DC5). The DHARD dither was at 23.3 Hz and the DSOFT dither at 21.1 Hz.

The new input matrix not only rejects hard mode much better, but also has a significantly improved SNR.

the interferometer locked by itself after we left this morning. It stayed locked for 10 hours (until the 6.4 EQ in Tonga a couple hours ago).

Some troubles re-locking:

1. arms locked green sort of OK although the ETMs are still swinging too much and there is lots of HOM green locks. The ALS unlocks drive up the ETMs; the damping settings ought to be made stronger for this part of the locking. ETMX in particular is kicked much more than ETMY.
2. ALS DIFF breaks the lock too much. Also there is something not right about ALS WFS scripting. The gain ramping or sequencing is off and often it comes on too slow to keep the y-arm locked.
3. sometimes there are Guardian thread exceptions related to channel access. I checked that these channels are responsive using 'caget' from terminal. Errors from log attached. Looking through  the logs its looks like this is happening on many of the attempts to go DOWN. IF this is not caught by a person, it could mean that the Guardian is not putting us back into the correct state for initial locking. Since the SDF is not up to date, sometimes our acquisition problems could be due to this. Perhaps we need some error flag (besides the error logs with a zillion lines of messages) if the Guardian fails to make its connections?

Rana, Dan, Georgia, Craig, Hang

We did some ASC modifications for high power:

1). dither loops

We now control both the dc locking points for PRM pointing and SOFT dofs using dithering loops, similar to what has been done at llo. For the soft loops we use ADS PIT/YAW 4 and 5 to servo them to the spot position on the ETMs. For PRM, we dither ITMY using PIT/YAW 3 and then feed back to PRM to control their dc pointing.

For the soft loops, in order to achieve true unconditional stability at low frequencies, we AC coupled the regular SOFT control signals (z at 0, pole at 0.1 Hz). This compensated the offloading to the top mass at 0.1 Hz so the regular SOFT loops is flat at low freqs. For the dithering part, we put a pole at 0.1 Hz which then combined with the integrator at the top mass to form a true 1/f integrator.

The dither loops are now turned on after engage ASC for full ifo.

2). Input matrix decoupling.

Since we have the DC pointing of the soft loops controlled with the dithering loops, we do not need to ensure the TR QPD combos we use is insensitive to TMS drift. Thus we modified the input matrices for some soft dofs.

Specifically, for CSOFT YAW, we changed the input matrix as

We also updated the QPD offsets so that when we first engage the soft loops (before the dithering loops being turned on) they could still converge to a decent location.

After changing this matrix, it seemed that we solved the 3 Hz oscillations showed up in common yaw loops.

The next issue we need to consider is the 0.4 Hz oscillation showing in DSOFT/DHARD/CSOFT PIT, due to either loop cross coupling or the dP/dtheta instability. The first step we tried to do was to decouple DHARD from DSOFT. Specifically, we wanted to update the QPD combos going to CSOFT PIT B path which damps the soft mode from 30 mHz to 1 Hz. Our old matrix sees the hard mode signal about as large as the soft mode. Based on some dithering lines Rana set up (starting from gps 1229590287 for more than 1k senconds), we find a new TR QPD combo that rejects the hard mode better

We wanted to try this new matrix but accidentally entered it to CSOFT P and caused a lockloss. We have not yet test this new combo again, and not sure if it would solve the 0.4 Hz oscillation.

Contrast Defect Ratio at 2 W = 8.4 +- 0.2 ppm

Contrast Defect is as follows: 



The electric fields associated with this equation are shown in the fourth attachment.

When DARM is locked on RF at 2 watts and the OMC is locked, we get about 11.8 mA, or 14.0 mW of power on the OMC DCDPs.  This means the DARM offset is around 29 pm according to this plot.
During this measurement, we move the DARM offset from 29 to -29 pm and see what happens.

Procedure:
1) Lock OMC while DARM is locked on RF
2) Zero the QPD offsets for OMC ASC by switching to OMC dither alignment, zeroing, and moving back to QPD control
3) Zero the OMC LSC dither control signal by offloading to PZT2, then stop the feedback, parking the OMC on the carrier fringe
4) Change DARM1 TRAMP to 120 seconds
5) Move DARM1 OFFSET through zero (in this case from 9e-5 cts to -9e-5 cts)
6) Re-engage OMC LSC by stopping input, clearing history, unholding output, and restarting input.  This will find the fringe again quickly for the next measurement.

Parameters during measurement:
            Input Power  = 1.89 W
     PRG at 0 pm Offset  = 46.8 +- 0.1
      DCPD Responsivity  = 0.858 A/W
DCPD Quantum Efficiency  = 0.98
   OMC DCPD Dark Offset  = 7.2e-5 +- 3e-5 mW  (Found at the last time the IMC was OFFLINE)
      OMC DCPD NULL/SUM  = 0.0060 +- 0.0015
   Measurement GPS Time  = 1229475480


Differences between 0 and 29 pm DARM offset:
Value         29 pm     0 pm    29 pm/0 pm
------------------------------------------
PRG            46.5     47.0         0.99
TRX [cts]    1502.    1518.          0.99
TRY [cts]    1555.    1572.          0.99

Assumed to be one:
- IMC Transmission
- Input Faraday Transmission
- Output Faraday Transmission
- OMC Transmission
- Mode Matching

Not yet considered:
- RF45 Transmission through the OMC
- Source of contrast (differential lens)

Dark Port alogs:
45753 DARM offset calibration
45734 OMC DCPD Sum/Null Matrix

Craig, D Brown, Georgia, Hang, Rana

Since last night, our adjustments to the SOFT loops has enabled stable operation at 25 W (> 1 hour) and several minutes of stability at 30 W.

• We dither the ETMs, and one ITM in both pitch and yaw to produce angle to length signals. We demod these to produce spot  position error signals on the mirrors and lock down these spots by adding them to the SOFT loop error points.
• Lowered the CSOFT yaw loop gain by a factor of 2 to minimize 3 Hz suspension oscillation.
• mix REFL WFS into CSOFT to minimize 3 Hz gain
• Sidles-Sigg RP compensation set to 25 W

The attached image shows the 30W unlock. The step from 25 W to 30 W happens at -180 seconds on the plot. The growing oscillation at the end is at 0.42 Hz as we've seen before.

Craig, Dan, Rana

- We lost lock from high power (24 Watts) at Dec 22 2018 09:36:422 UTC, and when we did an ezca connection error caused the guardian to stall and not complete the DOWN state.  Also, we had been using the Alignment Dither System (ADS) to align the arms, which was not turned off by the guardian even after STOP; EXEC-ing it.  This resulted in the four quad suspensions saturating.  Rana brought the quads to the SAFE state, but even in the SAFE state the tidal servo got to them and caused more saturations.
The guardian in PREP_FOR_LOCKING should now turn off all ADS CLKGAINs and feedback gains, and clear their history.

- We had another lockloss in OFFLOAD_DRMI_ASC, so this has not been solved by MICH ASC gain increases.  

- We had more LOCKING_ALS locklosses where DARM rings up while trying to acquire.  Characteristics of these locklosses include transmission dips in ALS-C_TRX where ETMX is being transitioned to DIFF control, and exponential DARM ringups at ~ 1 Hz.

- PRCL gain increases to 18 in ADJUST_POWER now.

- I turn on the LSC lines in DRMI_TO_POP now.  They should be turned off in PREP_FOR_LOCKING.

We commissioned some soft dithering loops for locking their DC positions to fixed spots on the ETMs. This is similar to what is used at LLO.

Specifically, after we increased power to 20 W, we used PIT4/PIT5 and YAW4/YAW5 to dither ETMX/Y PIT and YAW, respectively. We then demodulated the A2L response in DARM control and servoed the signal back to X/Y SOFT to set their DC locking point.

To make the control unconditionally stable, we turned off the FM3 (integrator) and FM4 (z=0.01Hz, p=0.1Hz) filters in the C/DSOFT control filter bank so that those filters approached DC as a constant (flat for < 0.1 Hz) . Meanwhile, the dithering loops are just simple integrators with 1/f shape in the freq band of interest. (Also there is another integrator in the top mass suspension). Such an 1/f - constant crossover guaranteed loop stabilities for small dithering loop gain. The crossover frequency is around a few mHz for now, which seemed fine.

By engaging the dithering loop the A2L noise is also improved significantly.

==================================================

More details:

The dithering frequencies are

 The control filter setups are attached to this log. Note that we changed the BP filters in DEMOD_SIG to FM6 (though many of them had the same names as the original BP filters), which was a 4th order cheby1 filters with width of 0.06 Hz.

We also created a guardian state SOFT_DITHER for engaging the dithering loops. However, we have not yet connected this node to the main guardian and will test it tomorrow.

This is another look at the ASC sensing matrix^$ that Hang and I have been looking at.

This attached plot shows the sensing matrix coefficients as a function of time. For this measurement, each of the 4 DoFs was driven at a different  frequency in YAW. We then demod the ASC sensors as a function of time to see how the sensing matrix evolves. Some notes:

1. The sensor data are scaled so that the line heights are similar. i.e., its as if some extra digital gain has been applied, but the same scaling is there for all 4 plots.
2. at t=5, the power has just been increased to 20 W. The input power goes to 23 W and t = 30 min.
3. The I & Q phases here refer to the audio demodulation, so the changing Q/I ratio is describing a change in the opto-mechanical response phase. This is NOT the same as a change in RF phase or Gouy phase.
4. at 31 minutes in this plot we tried adjusting some input matrix to remove cHard from cSoft.
5. at 47 minutes, the input power is increased from 20 to 23 W

the main effect here is that the cHard and cSoft couple more strongly into the PRC dOfs as the heating happens. So the matrix inversion will be power dependent as we were worried about. Might be that we'll have to do some SPOP norm to get it to be stable.

[$]: Reminder - the ASC sensing matrix is the thing that relates mirror angles (either in the mirror basis, or some interfero basis like chard, cSoft, MICH, SRC,..) to the signals read out by the angle sensors (WFS, optical levers, DC QPDs, etc).

Saravanan and Shivaraj

Since the installation of new time-dependent calibration parameters tracking model, we weren't able get some reasonable numbers for the tracking parameters (known as kappas) (a-log 45766). One of the main reasons was that we didn't we have good lock stretches with stable IFO configurations to test/debug the model. During ER13, there were a few such good lock stretches so we decided to look at one. Specifically we looked at GPS times 1229132118-600 when all the signal outputs from the calibration TDEP model were almost constant (so that we can concentrate on systematic). 

The first plot show the kappas as output by the front-end with then installed TDEP model. We see that kappas that track the actuations change are close what we expect (the signs of UIM and PUM stage were wrong though, they shouldn't be negative) but kappas that track the open loop IFO response are very different than what we expect. So we looked at the signal flow in the TDEP calibration model. To understand and compare what we would expect for signals at different point of TDEP model, we wrote a matlab code mimicking the signal flow. This code is similar to PCalmon we used to run during O1 and O2, but this code's output is continuous as front-end and uses the same filters as front-end (loaded from front-end model). This way we can calculate what we expect for signals at different points of TDEP model and compare against what we see. Looking at this code's output and TDEP model we found that the final complex blocks in TDEP model for IFO response (coupled cavity and spring) were supposed to be complex division but the installed ones are complex multiplication. Second figure show the MEDM and simulink model with the block circled for coupled cavity part. Same thing applies for the optical spring part.

The advantage of the matlab script is that we can regenerate the data with any changes. The last figure show what we would have seen if the last block was correctly installed as complex division. We see that the IFO response part would have come out as okay-ish. However the sign of optical gain is negative which is not correct and the optical spring parameters are not quite what we expect. If we change the sign of EPICS parameters EP2, EP3, EP4 installed via a-log 45766, then the actuation and coupled cavity pole parameters all come out okay (proper sign and magnitude similar to what we expected). But the spring parameter values still stay the same i.e., bit off then what we expect. We will try to further debug this, but at this point it would be better to do a full set of calibration measurements in a good low noise stable to build a new model. Since the UIM and PUM stage kappa values are nearly 10 % off of the initial model values, may be something not quite right in the current model (during the whole of O2, the kappa_pum hardly changed by a percent which is what we expect). Also at this point it would be good to move the lines to its final configuration.

The matlab script used for the comparison is added to svn at,

aligocalibration/trunk/Common/MatlabTools/calcsTools/calcsTDEPMatlab.m

(Kyle R, Gerardo M)

Uncrated the recently received large ion pump at X-Mid, discovered that each of the high voltage feedthroughs at the pump had one of the two holes plugged with a half broken screw still stuck on one its threads, fortunately it was the bottom hole for both of the feedthroughs, which makes for an easy extraction.  After both holes were successfully cleared using pliers, hammers and drill bits, the pump was tested.  The large ion pump passed its tests, both sides tested good.  The ion pump is ready to be installed next year.  Tested SN70105.

Note: to be able to repair the high voltage fastening holes on the ion pump body, we had to remove the pump from the wooden pallet/crate, we used the crane to lift it and place it on a block, the crane activity started at 22:52 UTC.

These were shut down earlier today.  T

This setup was used last week for demonstration and initial "shake down" testing of the first articles of our next generation Main Turbo pumping stations that will replace our existing "legacy" units over the course of the next year(s)?  They were not connected to the beam tube but, rather, were operated in a free standing configuration.