Stefan, Evan

Since we turned off the TCS last night and left PRMI locked on carrier, we have roughly 8 hours of good data that tells us the (1) the contrast defect and (2) the behavior of the ITM thermal lens as a function of time. To extract these quantities, we did the following:

• We extrated minute trends of LSC-POP_A_LF_OUT, LSC-REFL_A_LF_OUT, and LSC-ASAIR_A_LF_OUT from last night.
• We calibrated POP_A power to PRC power as follows:
  • FM10 already calibrates the POP_A counts into microwatts of power incident on the PD.
  • In Kiwamu's recycling gain measurement (LHO#15793), he misaligned PRM and locked MICH on a dark fringe. There was 7.2 μW on POP_A. Assuming a PSL power of 10.7 W (measured today), an IMC transmission of 83 % (LHO#9954), and a PRM transmission of 3.0 % (galaxy), this means there was 260 mW of power in the PRC.
  • The calibration PRC / POP_A is therefore 3.7×10⁴ W/W.
• We calibrated ASAIR_A power to SRC power as follows. Note that what I call "SRC power" is power leaving the BS going toward SRM (this is to distinguish it from "AS power" seen after the SRM).
  • ASAIR_A is already calibrated into microwatts of power incident on the PD.
  • In Stefan's contrast defect measurement (LHO#15332), he found the maximum SRC power from the free-swinging michelson was 1890 ct on ASAIR_B (not A). Today I measured the calibration to A is ASAIR_A / ASAIR_B = 3.52 ct/ct. The maximum SRC power from the free-swinging Michelson I take to be 10.7 W × 0.83 × 0.03 × 0.986 (this last number is the ITM reflectivity).
  • The calibration SRC / ASAIR_A is therefore 39 W/W.
• We calibrated REFL_A to REFL (i.e., the power heading from the PRM to IM4) as follows:
  • FM10 already calibrates the REFL_A counts into milliwatts of power incident on the PD.
  • In yesterday's RFAM measurement (LHO#15781), we measured 81 mW on REFL_A with the PRM aligned and the ITMs misaligned. The REFL power I take to be 10.7 W × 0.83 × 0.97.
  • The calibration REFL / REFL_A is therefore 106 W/W.
• It can be shown that the ratio P_SRC / P_PRC = T_M R_I, where TM is the carrier power transmissivity of the Michelson, and R_I is the power reflectivity of each ITM. I assume R_I = 0.986. Therefore, we can solve for T_M as a function of time given the calibrated power measurements. This is given in the upper plot of the attachment. From last night's data alone, the starting value of T_M is 1.115×10⁻³. However, Daniel has previously corrected the contrast defect for the presence of the 45 MHz sidebands (LHO#15432), finding a defect of 0.99×10⁻³. In the plot, I have rescaled T_M by 0.99 / 1.115 to reflect this.
• To convert ths into a change into diopters according to Stefan's formula, we need to know the "best" contrast defect achievable (i.e., the residual defect assuming no RoC mismatch). Stefan's measurement (LHO#15332) showed a minimum in ASAIR_B of (−0.2 + 1.78) ct with the Michelson locked on a dark fringe. According to Daniel's analysis, 0.17 ct of this is due to the 45 MHz sideband. So the net minimum is 1.41 ct. With a bright fringe power of 1890 ct, I take the "best" contrast defect to be 8×10⁻⁴.
• By using this "best" defect and Stefan's formula above, we compute change in the ITM thermal lens (in diopters) as a function of time. This is given in the lower plot of the attachment.

In principle, we can also use the above data to extract the mode-matching into the PRMI as a function of time. Perhaps we will pursue this later.

The carrier recycling gain was measured to be 45 with the three ASC loops closed in PRMI.

Please forget the previous measured recycling gain (alog 15527).

(Some numbers)

• POP_A increased from 7.2 uW (which I subtracted a dark offset of -6.3 uW) to 10350 uW
• According to the galaxy web page, I assumed Tp to be 3.1 %.
• Therefore Gp = 10350 uW / 7.2 uW x Tp =  45

• Also, for cross-check, I looked at POPAIR_A as well.
• POPAIR_A increased from 29.6 uW (with an offset of -0.8 uW being subtracted) to 43200 uW.
• This gives a recycling gain of 43500 uW / 29.6 uW x Tp = 45

Note that the IMC incident power was at 10 W during the measurement. REFL_LF dropped from a nominal of 83 mW to 6.3 mW when the PRMI was locked. The dark port ASAIR_A_LF stayed at 12000 counts during the lock. We could see a donuts mode at the dark port digital camera.

We shut off the ITMX CO2 laser at 20:00:40 in local time (4:00:40 UTC) for tomorrow's HWS project. We are leaving the PRMI locked on the carrier to see what happenes.

The lock held for about 8 hours, from about 8 PM to 4 AM local time.

A separate comment:  Hugh and I just checked which ground instruments are used where. It looks like HAM2 uses intrument A, and HAM3 uses instrument B.  We probably want those two chambers to use the same sensors. We might want to investigate a bit more on HAM 3 (and fix it!)  before we swap.

Here are coherences at the OUT of the Match filter.  Results look somewhat similar but with cetain differences to the plots Fabrice put in.  In particular, the X & Z coherences don't go as low in frequency.  The Y dof does look very similar.

Distinguishing glitch and operator initiated actions in PIT and YAW signals:

  We  can distinguish the glitch and operator actions by looking at their spectral signatures.   A glitch would cause a rise in spectral amplitude right across the entire frequency range.  This would then appear as a white line running vertically (across all frequencies) in the spectrogram.  Where as an operator initiated action would have a subsequent suspension damping motion at low frequencies (only). 

   We can see examples of both in the PIT spectrogram.  There are no glitches in the red trace (the spectrogram for that is in bottom panel). This was after about 7PM and folks had already started using the BS oplev for damping.  So their initial alignment efforts show up as small steps with an associated low frequency spectral signature. 

    The blue trace has the classic glitch related signals showing up in pitch.  They can be seen starting at 1.3 hrs and going on till 1.4 hrs.  I dont think anyone was using the IFO at that time.  Since the BS oplev is used for local damping continuously, it is likely that the gliches kicked the optic and caused the activity we see around that time.

  The picture is more messy in the case of YAW as we can see from the blue trace and its associated spectrogram (middle panel).  The yaw signal seems to be continuously affected by the glitching however the event we saw in pitch at 1.3 hrs can also be seen in yaw.  Once again there is no operator related activity in the blue trace while the red trace shows some steps which have an associated low frequency spectral signature (bottom panel).  I concluded that they were associated with the initial alignment activity which was going on at that time.

I looked at whether the improvement in the laser quality has resulted in an actual improvement in the BS local damping.  There is a tangible improvement in YAW.

1) The Spectrogram of YAW motion shows that the injection of broadband noise into the optic motion in YAW due to glitching has disappeared after the swapping of lasers

2) the Coherence between the witness channel and Oplev channel in YAW shows that we can now extend the servo bandwidth to about 10Hz reliably.

3) The spectrum of yaw motion dropped by a factor of two in the range 1 to 20 Hz. This probably has nothing to do with the laser per se.  Probably the pier motion decreased between the two data segments.

Performance check after a week of operation

    To see if the laser is still operating safely within the glitch free region, I checked the 1s trend over the past two days.  The laser power has a slow drift of about 1% in a day.  This is probably a LVEA average temperature related effect.   The long term spectrum shows a 1/f shape down to 10^-4 Hz.

And to see the broad band noise I looked at raw signal over the past four hours (256 samples/sec)

The 4hr stretch of raw data spans a period when the oplevs were not used for first 1.4 hour stretch and then were turned on.  We can see the suspension resonances damp in the witness channels.  

The spectrograms show that there is broad band noise in the optic motion, but it is not due to the laser glitching. 

The top panel shows the laser spectrogram and it does not show any broadband noise.

Conclusion:

     The laser is performing well, without glitches.   All the action we see in the Pitch and Yaw is associated with either human intervention or lock loss events which have kicked the optic.

After looking at the oplev spectra with the OL damping loops on and off, I turned down the yaw gain from 650 ct/ct to 500 ct/ct to reduce the amount of extra noise injected between 1 and 10 Hz. The pitch gain is still 300 ct/ct.

In the attached plot, blue is the spectrum without damping, and red is the spectrum with the new damping gain.

We locked the PRMI on carrier. The carrier recyclying gain was measured to be 35 at highest. However, since the alignment was not perfect, it probably would go up. To be continued.

After playing with the gain settings, we eventually became able to lock the PRMI on carrier. However the alignment was not stable to keep it locked with high build-up. I think this needs more study to understand what is going on. Anyway, so far, the highest buidl-up in POPAIR_A_LF we had tonight was about 3.5x10⁴ uW. When the simple MICH without power-recycling was locked, POPAIR_A_LF was about 30 uW. Assuming that there is no mode-mismatch and Tp=0.03, we get a recycling gain of 3.5x10⁴ / 30 * Tp = 35. 

LSC settings:

• set LSC_CONFIGS to PRMI_SB_OFFLOADED
• flip the PRCL control sign so that the PRCL_GAIN is -11
• switch the sensor for MICH from REFL45 to AS45Q with an input matrix element of 1.
• set LSC-MICH_GAIN to -100.
• flip the sign of POPAIR_B_RF18_I in the trigger matrix so that it locks when RF18 is negative.

Attached are the OMC scan results for a PRMI sideband lock, compared to a scan from a single-bounce beam.  The first plot shows the results of three single-bounce scans and three PRMi scans (100 second ramps of PZT2); the second plot has averaged the traces.  The PRMI data appears to be shifted upwards compared to the single-bounce data, by about 1V in the PZT2 output.  We expect some drift and hysteresis in the PZT, but the single-bounce data was taken immediately after the PRMI lock, and a shift of this size is...surprising.

Using Kiwamu's expression from alog:14532, I calculate the PRC gain of the 45MHz sideband to be about 11.8 or 14, depending on which sideband peak you use.  I think this is lower than we expect.  The PRMI sideband lock was quite wobbly with a lot of angular motion, we might get a more robust measurement by locking the OMC to a particular mode and maximizing the transmission.

Here is a table of peak heights:

With a Schnupp asymmetry of 9.5cm the gain, for example for the upper 45MHz sideband, is (4.7/0.31) * (0.03*0.5*0.5) * (1/sin(2*pi*0.095*5*9100230/c)**2) = 13.9.

Just for a book-keeping purpose:

Two weeks later from this entry, we have measured the recycling of the carrier with the ASC loops fully engaged. We measured it to be 45 (see alog 15793).