Measurement of Schnupp asymmetry

This was done by measuring the relative phase between the sidebands reflected from the two arms while the arm cavities are locked.

The Schnupp asymmetry is measured to be:   L_sa = 3.64 ± 0.32 cm

Description:

As a phase reference we use the zero crossing of the response function for the out-of-phase control signal for the single arm cavity lock [0]. The difference in the RD rotation phase of the response zero crossings indicates the phase difference in the sideband signals reflected from the arms. Assuming the asymmetry is less than half the RF modulation wavelength [1], the asymmetry is given by the following formula:

       \Delta \phi   c   1 
L_sa = ----------- ----- -
           360     f_RSB 2

We use a LSC digital lock-in to measure the response of the arm cavity at a single-frequency drive of it's end mirror.

[0] The locations of the zero crossings in the out-of-phase components of the response can be determined to higher precision than the maxima of the in-phase components.

[1] f_RSB = 55 MHz,     c/f_RSB/2 = 2.725 m

Procedure:

1. Lock/tune the Y arm only.
   
   • We use AS55_I to lock the arms.
2. Engage the LSC lock-in.
3. Tune the lock-in parameters:

lock-in freq: 103.1313 Hz
I/Q filters:  0.1 Hz low-pass
phase:        0 degrees

4. Set as input to the lock-in the out-of-phase quadrature from the control RFPD.  In this case AS55_Q->LOCKIN.
5. Drive the arm cavity end mirror by setting the LOCKIN->Y_arm element in the control matrix.
6. Note the "RD Rotation" phase between the demodulated signals from the control PD (AS55)
7. For some reasonable distribution of phases around the nominal "RD Rotation" value, measure the amplitude of the lock-in I output.
   • Assuming the Q output is nearly zero, it can be neglected.  In this case the Q amplitude was more than a factor of 10 less than the I amplitude.
   • Here we take 5 measurements, each separated by one over the measurement bandwidth (as determined by the lock-in low pass filter), in this case 10 seconds.  The figure above plots the mean of these measurements, and the error bars indicate the standard deviation.

The data and python data-taking and plotting scripts are attached.

Error Analysis:

To to determine the parameters of the response (which we know to be linear) we use a weighted linear least-squares fit to the data:

y = b X

where:

X_0j = 1
X_1j = x_j              # the measurement points
y = y_i                 # the response
b = (b₀, b₁)     # line parameters

The weighting is given by the inverse of the measurement covariance matrix. Since we assume the measurements are independent, the matrix is diagonal and W_ii = 1/\sigma_i² The
estimated parameter values are given by:

\beta  =  ( X^T W X )^-1 X^T W y  =  ( X'^T X' )^-1 X'^T y'

where X' = w X, y' = w y and w_ii = \sqrt{W_ii}.

The X' and y' are calculated from the data and passed into the lstsq routine. The output is \beta.

The error on the parameters is described by the covariance matrix M^\beta:

M^\beta = ( X^T W X)^-1 = ( X'^T X')^-1

with correlation coefficients \rho_ij = M^\beta_ij / \sigma_i / \sigma_j.

The x-axis crossing is then given by:

X(Y=0) = - \beta₁ / \beta₀

References:

Valera's LLO measurement
http://en.wikipedia.org/wiki/Weighted_least_squares
http://en.wikipedia.org/wiki/Linear_least_squares_(mathematics)#Weighted_linear_least_squares
http://en.wikipedia.org/wiki/Error_propagation

Some updates of the LSC screen

- Signal amplitude monitor for the PD signals (--> glows red for more than 1000)

- Kissel Buttons for the main matrices

- Trigger display at the output of the DOF filters

- Signal amplitude monitor for the SUS LSC output (--> glows red for more than 10000)

ADC Over flow monitor is showing some unknown numbers (as ADCs are handled by IOPs).
I asked Joe for the investigation (and consideration for the policies)

[Suresh, Koji]

   I used a matlab code written by Koji to analyse the transimpedance and current noise data  of REFL55.  The details are in the attached pdf file.

Resonance is at 55.28 MHz:

Q of 4.5, Transimpedance of 615 Ohms

shot noise intercept current = 1.59 mA

current noise =21 pA/rtHz

Notch at 110.78 MHz:

Q of 54.8 Transimpedance of 14.68 Ohms.

Last night I was making a script which will measure the sensing matrix using the realtime LOCKIN module.

The script is a kind of expansion of Jamie's one, which measure the asymmetry, to more generic purpose.

It will shake a suspended optic of interest and measure the response of each sensor by observing the demodulated I and Q signals from the LOCKIN module.

I will continue working on this.

  (current status)

 - made a function that drives the LOCKIN oscillator and get the data from the I and Q outputs.

 - checked the function with the MICH configuration.

   ITMX, ITMY and BS were shaken at 100 Hz and at different time.

   Then the response of AS55_Q showed agreement with what I got before for the actuator calibration (see this entry).

   It means the function is working fine.

I am now measuring the sensing matrix in the DRMI configuration.

A goal of tonight is to measure the sensing matrix as a test of the script.

The result will be updated later.

The sensing matrix was measured in the DRMI configuration for the first time.

The measurement was done by an automatic script and the realtime LOCKIN module built in the c1lsc model.

The resultant matrix is still too primitive, so I will do some further analysis.

(Measurement of sensing matrix)

 The quantities we want to measure are the transfer functions (TFs) from displacement (or change in optical phase) of each DOF to sensors in unit of [counts/m].

So essentially the measurement I did is the same as the usual TF measurement. The difference is that this measurement only takes TFs at a certain frequency, in this case 283 Hz.

 The measurement goes in the following order :

  (1) Lock DRMI

  (2) Shake an optic of interest longitudinally with an amplitude of 1000 counts at 283.103 Hz, where no prominent noise structures are present in any spectra of the sensor signals.

  (3) Put a notch filter at the same frequency of 283.103 Hz in each DOF (MICH, PRC and SRC) to avoid unwanted suppression due to the control loops.

       (This technique is essentially the same as this one, but this time the control loops are shut off only at a specific frequency )

       The notch filter I put has a depth of 60 dB and Q of 20. The filter eats the phase of ~10 deg at 200 Hz, which still allow servos to run with a high UGF up to 200Hz.

  (4) Take the output signal from a signal port of interest (i.e. REFL11_I, etc.,) and then put it into the realtime LOCKIN module.

  (5) Measure the resultant I and Q signals coming out from the LOCKIN module.

  (6) Repeat the procedure from (2) through (5) for each optic and sensor.

(Results)

 Again, the resultant sensing matrix is still primitive, for example the optic-basis should be converted into the DOF basis.

The values listed in the matrix below is the absolute values obtained by operation of sqrt( I^2 + Q^2) plus the polarity according to the output from I and Q of LOCKIN.

Therefore they still contain the actuator response, which is not desired. i will calibrate them into [counts/m] later by using the calibration factor of the actuator responses.

All the raw data showed the relative phase between I and Q either ~ 127 deg or ~ -53 deg.

In my definition, the one has 127 deg is plus polarity and the one has -53 deg is minus polarity.

Technically speaking the polarity depends on the polarity of the actuator and also the direction of the actuator against the DOFs.

Without any excitation the absolute values fluctuated at about 10^-4 - 10^-5, so the excitation amplitude was big enough to observe the sensing matrix.

Though, I still need to estimate the statistical errors to make sure the SNR is reasonably big.

  Fig.1 Measured sensing matrix from optic to sensors.

(Things to be done)

  - convert the optic-basis (i.e. BS, ITMs, PRM and SRM) to the DOF-basis (i.e. MICH, PRC and SRC) so that the matrix is understandable from point of view of the interferometer control.

  - estimate the optimum demodulation phase for each DOF at each sensor port.

  - add some statistical flavors (e.g. error estimations and so on.)

  - edit the script such that it will keep watching the ADC overflows and the coherence to make sure the measurement goes well.

  - add some more signal ports (e.g. REFL55, POY55 and etc.)

  - compare with an Optickle model

I was trying to measure the sensing matrix in the PRMI configuration, but basically gave up.

It is mainly because the lock of PRMI wasn't so stable and it didn't stay locked for more than a minute.

It looked like an angular motion fluctuated a lot around 1- 3 Hz. The beam spot on the AS camera moved a lot during the lock.

I have to figure out who is the bad suspension and why.

All the suspensions are bad until you fix them. But, ... there is a script which can be used to diagnose them today:

Python SUStest

I was able to measure the sensing matrix in the PRMI configuration.

The results will be posted later.

Here is the result of the measurement of the sensing matrix in the PRMI configuration.

If we believe the resultant matrix, it is somewhat different from what we expected from a finesse simulation (summary of simulated sensing matrix).

(Motivation)

As a part of the DRMI test plan, we wanted to check the sensing matrices and consequently diagonalize the LSC input matrix.

The matrix of the DRMI configuration has been measured (#4857), but it was a bit too complicated as a start point.

So first in order to make sure we are doing a right measurement, we moved onto a simpler configuration, that is PRMI.

(measurement)

The technique I used was the same as before (#4857) except for the fact that SRM wasn't included this time.

   - PRC was locked to the carrier resonant point. The UGF of MICH and PRC were ~ 110 Hz and 200 Hz respectively.

   - Longitudinally shook BS, ITMs and PRM at 283.103 Hz with an amplitude of 1000 counts using the LOCKIN oscillator in C1LSC.

   - Took the I and Q phase signals from the LOCKIN outputs.

The table below is the raw data obtained from this measurement :

(Conversion of matrix)

 With the matrix shown above, we should be able to obtain the sensing matrix which gives the relation between displacements in each DOF to each signal port.

The measured matrix connects two vectors, that is,

       (signal port vector) = [Measured raw matrix] (SUS actuation vector),   -- eq.(1)

where

       (signal port vector) = (AS55_I, AS55_Q, REFL11_I, REFL11_Q)T   in unit of [counts],

    (SUS actuation vector) = (BS, ITMX, ITMY, PRM)T   in units of [counts].

Now we break the SUS actuation vector into two components,

       (SUS actuation vector [counts])  = (actuator response matrix [m/counts])-1 * (MICH, PRM [m] )^T   -- eq.(2)

 where

       (actuator response matrix) =  2.05x10-13 * ( [1   ,  0.217, -0.216,   0  ],

                                                 [ 0.5,  0.109 -0.108, 0.862]  )  in unit of [m/counts]

These values are coming from the actuator calibration measurement.

In the bracket all the values are normalized such that BS has a response of 1 for MICH actuation.

Combining eq.(1) and (2) gives,

     (signal port vector) = (measured raw matrix) * (actuator response matrix)-1 * (MICH, PRM)T

And now we define the sensing matrix by

     (sensing matrix) = (measured raw matrix) * (actuator response matrix)-1

The sensing matrix must be 4x2 matrix.

For convenience I then converted the I and Q signals of each port into the absolute value and phase.

       ABS = sqrt((AAA_I)2 +(AAA_Q)2 ),

       PHASE = atan (AAA_Q / AAA_I),

where AAA is either AS55 or REFL11.

(Resultant matrix)

The table below is the resultant sensing matrix.

ABS represents the strength of the signals in unit of [cnts/m], and PHASE represents the demodulation phases in [deg].

There are several things which I noticed :

   - The demodulation phase of MICH=>AS55 and PRC=>REFL11 are close to 0 or 180 deg as we expected.

      This is a good sign that the measurement is not something crazy.

   - AS55 contains a big contribution from PRC with a separation angle of 152 deg in the demodulation phase.

     In AS55 the signal levels of MICH and PRC were the same order of magnitude but PRC is bigger by a factor of ~4.

     However the finesse simulation (see wiki page) shows a different separation angle of 57 deg and MICH is bigger by factor of ~6.

  - REFL11 is dominated by PRC. The PRC signal is bigger than MICH by a factor of ~100, which agrees with the finesse simulation.

    However the separation angle between PRC and MICH are different. The measurement said only 19 deg, but the simulation said ~ 90 deg.

  - Woops, I forgot to calibrate the outputs from the LOCKIN module.

    The whole values must be off by a certain factor due to the lack of the calibration , but fortunately it doesn't change the demodulation phases.

New LSC screen is 80% completed.

It is now accessible from the LSC menu of "sitemap".

Most of the part in the screen is clickable such that it launches another screen depending on the location of the click.

The bottom part of the screen still need some work.

RFPD screen is temporary

LSC control screen is also temporary

DAC overflow indicators are still broken.

Channel assignment of the whitening filters are arbitrary so far.

Of course I made a mistake in my calculation of the sensing matrix. I will figure out which point I mistook.

The MICH signal must have the demodulation phase of around 90 deg in AS55

because we had adjusted the demodulation phase such that the MICH signal mostly appears on AS55_Q.

LSC model has been updated and running,

- Now the power and signal recycling cavity lengths are named "PRCL" and "SRCL" in stead of three letter names without "L".

- Names for the trigger monitor were fixed. They are now "C1:LSC-DARM_TRIG_MON", etc., instead of "...NORM"

- Channel order of the DC signals for PDDC_MTRX and TRIG_MTRX were changed.

It was "TRX, TRY, REFL, AS, POP, POX, POY" but now "AS, REFL, POP, POX, POY, TRX, TRY".

We should change the locking script to accomodate these changes.

[Jenne, Koji]

We have tested the LSC whitening filters. In summary, they show the transfer functions mostly as expected (15Hz zerox2, 150Hz pole x2).
Only CH26 (related to the slow channel "C1:LSC-PD9_I2_WhiteGain. VAL NMS", which has PD10I label in MEDM) showed different
phase response. Could it be an anti aliasing filter bypassed???

The 32 transfer functions obtained will be fit and summarized by the ZPK parameters.

Method:

The CDS system was used in order to get the transfer functions
- For this purpose, three filter modules ("LSC-XXX_I", "LSC-XXX_Q", "LSC-XXX_DC") were added to c1lsc
in order to allow us to access to the unused ADC channels. Those filter modules have terminated outputs.
The model was built and installed. FB was restarted in order to accomodate the new channels.

- Borrow a channel from ETMY UL coil output mon. Drag the cable from the ETMY rack to the LSC analog rack.
- Use 7 BNC Ts to split the signal in to 8 SMA cables.
- Put those 8 signals into each whitening filter module.

- The excitation signal was injected to C1:SUS-ETMY_ULCOIL_EXC by AWGGUI.
- The transfer functions were measured by DTT.
- The excitation signal was filtered by the filter zpk([150;150],[15;15],1,"n")
   so that the whitened output get flat so as to ensure the S/N of the measurement.

- For the switching, we have connected the CONTEC Binary Output Test board to the BIO adapter module
   in stead of the flat cable from the BIO card. This allow us to switch the individual channels manually.

- The whitening filters of 7 channels were turned on, while the last one is left turned off.
- We believe that the transfer functions are flat and equivalent if the filters are turned off.
- Use the "off" channel as the reference and measure the transfer functions of the other channels.
- This removes the effect of the anti imaging filter at ETMY.

- Once the measurement of the 7 channels are done, switch the role of the channels and take the transfer function for the remaining one channel.

Result:

- We found the following channel assignment

• The ADC channels and the PDs. This was known and just a confirmation. 
• The ADC channels and the WF filter on MEDM (and name of the slow channel)

- We found that the binary IO cable at the back of the whitening filter module for ADC CH00-CH07 were not connected properly.
This was because the pins of the backplane connector were bent. We fixed the pins and the connector has been properly inserted.

- CH26 (related to the slow channel "C1:LSC-PD9_I2_WhiteGain. VAL NMS", which has PD10I label in MEDM) showed different
phase response from the others although the amplitude response is identical.

Summary of the channel assignment (THEY ARE OBSOLETE - SEPT 20, 2011)

ADC                    Whitening Filter
CH  PD                 name in medm   related slow channel name for gain
---------------------------------------------------------------------------
00  POY11I             PD1I           C1:LSC-ASPD1_I_WhiteGain. VAL NMS
01  POY11Q             PD1Q           C1:LSC-ASPD1_Q_WhiteGain. VAL NMS
02  POX11I             PD2I           C1:LSC-SPD1_I_WhiteGain. VAL NMS
03  POX11Q             PD2Q           C1:LSC-SPD1_Q_WhiteGain. VAL NMS
04  REFL11I            PD3I           C1:LSC-POB1_I_WhiteGain. VAL NMS
05  REFL11Q            PD3Q           C1:LSC-POB1_Q_WhiteGain. VAL NMS
06  AS11I              PD4I           C1:LSC-ASPD2_I_WhiteGain. VAL NMS
07  AS11Q              PD4Q           C1:LSC-ASPD2_Q_WhiteGain. VAL NMS
08  AS55I              AS55_I         C1:LSC-ASPD1DC_WhiteGain. VAL NMS
09  AS55Q              AS55_Q         C1:LSC-SPD1DC_WhiteGain. VAL NMS
10  REFL55I            PD3_DC         C1:LSC-POB1DC_WhiteGain. VAL NMS
11  REFL55Q            PD4_DC         C1:LSC-PD4DC_WhiteGain. VAL NMS
12  POP55I             PD5_DC         C1:LSC-PD5DC_WhiteGain. VAL NMS
13  POP55Q             PD7_DC         C1:LSC-PD7DC_WhiteGain. VAL NMS
14  REFL165I           PD9_DC         C1:LSC-PD9DC_WhiteGain. VAL NMS
15  REFL165Q           PD11_DC        C1:LSC-PD11DC_WhiteGain. VAL NMS
16  NC (named XXX_I)   PD5I           C1:LSC-SPD2_I_WhiteGain. VAL NMS
17  NC (named XXX_Q)   PD5Q           C1:LSC-SPD2_Q_WhiteGain. VAL NMS
18  AS165I             PD6I           C1:LSC-SPD3_I_WhiteGain. VAL NMS
19  AS165Q             PD6Q           C1:LSC-SPD3_Q_WhiteGain. VAL NMS
20  REFL33I            PD7I           C1:LSC-POB2_I_WhiteGain. VAL NMS
21  REFL33Q            PD7Q           C1:LSC-POB2_Q_WhiteGain. VAL NMS
22  POP22I             PD8I           C1:LSC-ASPD3_I_WhiteGain. VAL NMS
23  POP22Q             PD8Q           C1:LSC-ASPD3_Q_WhiteGain. VAL NMS
24  POP110I            PD9I           C1:LSC-PD9_I1_WhiteGain. VAL NMS
25  POP110Q            PD9Q           C1:LSC-PD9_Q1_WhiteGain. VAL NMS
26  NC (named XXX_DC)  PD10I          C1:LSC-PD9_I2_WhiteGain. VAL NMS
27  POPDC              PD10Q          C1:LSC-PD9_Q2_WhiteGain. VAL NMS
28  POYDC              PD11I          C1:LSC-PD11_I_WhiteGain. VAL NMS
29  POXDC              PD11Q          C1:LSC-PD11_Q_WhiteGain. VAL NMS
30  REFLDC             PD12I          C1:LSC-PD12_I_WhiteGain. VAL NMS
31  ASDC               ASDC           C1:LSC-PD12_Q_WhiteGain. VAL NMS
---------------------------------------------------------------------------

[Jenne / Kiwamu]

 Last night we modified the locking scripts, that were called from C1IFO_CONFIGURE.adl, to adapt them to the new "PRCL" and "SRCL" convention.

So far they work fine and quitted dumping some error messages about inexistence of these channel names.

  P.S. The locking scripts have been summarized on the 40m wiki

I have fit all of the LSC whitening filters using vectfit4.m

All the data is in my folder ..../users/jenne/LSC_WhiteningTest_29June2011/

The zpk info is saved with each plot of the fit.  The pdfs are kind of huge to stitch together (or rather my computer doesn't want to do it), so I'll just post a representative one for now.

During the daytime either tomorrow or Friday I'll adjust the actual dewhitening filters to match the measured zpk values.

 I made a handy-dandy table showing the zpk values for each whitening filter in the wiki: New whitening filter page

Next on the whitening filter to-do list: actually put these values into the dewhitening filters in foton.

The LSC model has been updated:

Binary outputs to control whitening filter switching

We now take the filter state bit from the first filter bank in all RF PD I/Q filter banks (AS55_I, REFL11_Q, etc) as the controls for the binary analog whitening switching on the RF PD I/Q inputs. The RF_PD part was also modified to output this control bit. The bits from the individual PDs are then combined into the various words that are written to the Contec BO part.

Channel mapping updated/fixed to reflect wiring specification

Yesterday Suresh posted an updated LSC wiring diagram, with correct channel assignments for the RF PD I/Q and DC inputs.  Upon inspection of the physical hardware we found that some of LSC the wiring was incorrect, with I/Q channels swapped, and some of the PDs in the wrong channels.  We went through and fixed the physical wiring to reflect the diagram.  This almost certainly will affect the EPICS settings for some of the input channels, such as offsets and RD rotations.  We should therefore go through all of the RF inputs and make sure everything is kosher.

I also fixed all of the wiring in the LSC model to also reflect the diagram.

Once this was all done, I rebuilt and restarted the LSC model, and confirmed that the anti-whitening filter banks in the PD input filter modules were indeed switching the correct bits.  I'll next put together a script to confirm that the LSC PD whitening is switching as it should.

[Jamie, Jenne]

We decided to take on the deceptively easy-sounding task of checking that the LSC whitening switching was happening as anticipated.  We hoped to discover that when we clicked the "unwhitening" switches in FM1 of the LSC PDs, we would see the analog whitening turn on and off for the matching channel.  That is what is supposed to happen.

Tragically, it is instead one big giant crazy disaster of a mess.

What we did:

Made a 24tapus (octopus like last time, except more...), with a 50kOhm resistor as our white noise source (instead of using a DAC channel and AWG). 

We plugged our 24tapus into the 3 of 4 whitening boards on the LSC rack that are currently in use.  One of the boards just has 8 terminators on the input, so we left that one alone for now. 

We put the whitening gains to 0dB so that all the channels looked the same. 

We looked at the PD _IN1 channels in DTT, and monitored which signals had whitening switching when we clicked the "unwhitening" buttons on the PD filter banks. 

So far, we can find no rhyme or reason as to why some of the channels work (click unwhite on that PD, see that signal have whitening switching), and others don't.  Some channels we just can't get to switch no matter what, others are just mis-mapped.  There is no discernible pattern.

What we think (so far) is going on:

All of the cables from the PD demod boards are going to the Whitening board inputs, exactly as in Suresh's Diagram.  The only difference is that Refl33, AS165 and Refl165 demod boards don't exist in the rack at this time. 

The Whitening and AA boards in Suresh's Diagram labeled 0-7 are connected to Binary Output channels 0-7. This is a good thing.

The Whitening and AA boards in the diagram labeled 8-15 are connected to Binary Output channels 24-31. This is not so awesome.

This is all we are confident about at this time.

Next steps:

We are hoping that Ben has a secret stash (or can tell us who would) of LSC rack wiring diagrams.  We would like to find out, without the pain of tracing wires and cables by hand, how the Binary I/O information gets through the cross-connect on the LSC rack up to the whitening boards. 

We are leaving the 24tapus in place for now, so that we can carry on tomorrow, either with a wiring diagram in hand, or carefully tracing cables. 

Remember, as per our marker board conversation, that the resistor noise as excitation method only works if the gain of all of the channels is set to something high (like 45 dB).

At 0 dB, the resistor noise is only 30 nV/rHz, whereas the ADC noise is more like 10000 nV/rHz...

Yesterday we started going through the LSC binary whitening switching to make sure the new switching control in the LSC model was working.  Jenne and I hooked up a fancy home-brew white noise generator [0] into all of the LSC whitening filter inputs and started switching the whitening filters to see what would happen.  We found that some of the channels were switching, but the majority were not, or worse yet switching the wrong channels.  Upon closer inspection, and after finally finding the LSC wiring schematic, we found that the LSC rack cross-connect/back-plane cabling was pretty much a complete mess, and didn't at all correspond to the channel layout in Suresh's diagram.

Given that the back-plane wiring had to be almost completely redone, we decided to completely redo the LSC electronics layout, to be a little more self-consistent and to use the given space more efficiently.  We'll post an updated electronics drawing soon.  The LSC model was also updated to reflect the new layout.

We then went through and verified that all of the whitening switching was working with the new layout.  As described previously, the first filter in the PD input filter bank is used to control the switching.  We did indeed verify that all the switching is working, but we noticed that switching logic was inverted, such that the whitening filter engaged when the filter was turned off.  This was fixed in the model and all the switching logic was verified to be working as expected.

Everything has now been hooked back up, but we need to verify that we're getting all of the PD demodulated RF and DC outputs as expected.  We need to check the RF phases, as some of the RF cable lengths have changed.

[0] a 50k resistor

Links:

• LSC wiring schematic (linked from Jay's homepage)

The ABSL locking setup to the PSL is down. 

According the plan, I started to use the IR beam dumped after the doubling crystal for the IR beat lock (Sonali's project).  The beat lock was disturbed when I shifted some clamps to make way for a few mirrors.  So I set about fixing the beat lock.  I reobtained the lock but noticed that the net beam power reaching the Newfocus 1611 detector was around 15mW.  10mW from the ABSL and 5mW from PSL.

This is much too high as the maximum allowed on 1611 is 2mW. 

I therefore started to adjust the power levels by using  Y1-1064-45S mirrors at non-45 deg angles.  However Rana pointed out that this would lead to amplitude noise due to the mirror vibrations.  I then switched to using beam splitters as pick offs.   This is better than using neutral density filters since the back scatter is lower this way.

David wanted some of the ABSL beam for his SURF student.  So I changed the mirror after beam expanding telescope on the ABSL route to provide this power.  We also installed a pair of half wave plates and a PBS to allow us smooth power level control on this beam.

The beat lock setup is now down and needs to be completed for PRCL and SRCL measurements.

Through some locking exercise I found that several things are degrading.

Remember the interferometer is like a cat, so we have to feed and take care of her everyday. (Otherwise the cat will be dead !)

Beam axis:

Locking of the Arms :

Locking of PRM :

 I had to realign PSL beam into the MC in order to reobtain the MC lock.  We lost lock at sometime around 8:30 AM on Tuesday.  See attached trend data for MC_RFPD_DCMON. 

The is the second time this week that I had to do this when we were unable to obtain the MC lock.  On both occassions the zig-zag at the end of the PSL table was tweaked to minimise the MC_RFPD_DCMON.

 We have been using the MC as a Beam Axis Reference.  And therefore we are adjusting the PSL beam to maximise coupling into MC.  However if MC's beam axis has shifted, then would it not be best to use the pzt's to re-obtain coupling into the arm cavities? 

[Nicole / Jamie / Rana / Kiwamu]

  The X arm and Y arm have been locked.

The settings for the locking were stored on the usual IFO_CONFIGURE scripts, so anybody can lock the arms.

In addition to that Nicole, Jamie and Rana re-centered the beam spot on the ETMY_TRANS camera and the TRY PD.

The next step is to activate the C1ASS servo and align the both arms and beam axis.

Xarm locking notes:

* Changed TRX gain from -1 to -0.02. Without this 50x reduction the arm power was not normalized.

* Had to fix trigger matrix to use TRX for XARM and TRY for YARM. Before it was crazy and senseless.

* Lots of PZT alignment. It was off by lots.

* Yarm trans beam was clipping on the steering mirrors. Re-aligned. Needs to be iterated again. Be careful when bumping around the ETMY table.

* YARM gain was set to -2 instead of -0.2. Because the gain was too high the alignment didn't work right.

ALWAYS HAVE an OPEN DATAVIEWER with the standard ARM channels going when doing ANY INTERFEROMETER WORK.

THIS IS THE LAW.

I modified the script armloss so that the channel names in the script are properly adopted to the new CDS.

Additionally I disabled the ETMX(Y)_tickle command in the script.

The tickle command puts some offsets on the LSC signal to let the arms pass through a fringe until it gets locked, but apparently we don't need it because the arms are loud enough.

A brief check showed that the script ran fine.

I will measure the loss on the X and Y arm cavity tomorrow.

 Required arm length = 37.7974 +/- 0.02 [m]

This is a preliminary result of the estimation of the Arm length tolerance.

Figure.2  A sensing matrix of the 40-m DRFPMI while changing the position of ETMX/Y by \sigma = 2 cm.

I did the measurement of the arm loss on both X and Y arm by running the armLoss script.

The results will be posted later.

wow. This Monte-Carlo matrix is one of the most advanced optical modeling things I have ever seen. We never had this for any of the interferometers before.

Old MZ PD (InGaAs 2mm, @29.5MHz) has been modified for REFL33.
There has been no choice for the 11MHz notch other than putting it on the RF preamp
as the notch in parellel to the diode eats the RF transimpedance at 33MHz.

I wait for judgement of Rana if the notch at the MAX4107 feedback is acceptable or not.

So far, this is just preliminary, because I haven't done full error analysis to determine the error on my measurements.  That will hopefully be done by tomorrow afternoon (so before we start taking off doors).

I find that the length of the Xarm is:  37.5918 meters.

I find that the length of the Yarm is:   37.5425 meters.

I used the mass-kicking technique, as summarized by Kiwamu, and fully described by Alberto.  More words / description to follow with the full error analysis.

REFL165 PD has been made from the old 166MHz PD.
As the required inductance was ~10nF level, the stray inductance of the circuit pattern was significant.
So, I am not so confident with the circuit functionality before the optical transfer function test.

I will test REFL33 and REFL165 with the Jenne laser to see how they work.

This REFL165 was good in terms of RF, but I forgot to make the DC path functioning.

I will try some ideas to fix this tomorrow.

REFL33 is ready for the installation

Characterization results of REFL33 is found in the PDF attachment.

Resonance at 33.18MHz
Q of 6.0, transimpedance 2.14kOhm
shotnoise intercept current = 0.52mA (i.e. current noise of 13pA/rtHz)

Notch at 10.97MHz
Q of 22.34, transimpedance 16.2 Ohm

Notch at 55.60MHz
Q of 42.45, transimpedance 33.5 Ohm

REFL165 PD was made and tested. The characterization results are in the PDF file.

Resonance at 166.12MHz
Q of 7.3, transimpedance 667Ohm (Series Resistance = Z/Q² = 2.5Ohm)
shotnoise intercept current = 4.3mA (i.e. current noise of 36pA/rtHz) 

As the circuit pattern had ~10nH level strain inductance, some technique was needed.

• The diode was pushed in so as to reduce the lengths of the legs as short as possible.
• The inductor for the resonant circuit has been located as close to the photodiode as possible
• The other side of the inductor was needed to be bypassed by a large (0.1uF) capacitor, as the original circuit pattern (D1-L5-C33//R22) was too skinny and long.
• C32 is also moved next to the diode.
• The path of the photo current circuit was made thicker by Cu tapes.

Now the size of the loop for the resonant circuit is comparable with the size of SOIC-8 opamp.
(Left-Top corner of the photo)

This improved the resonant gain by factor of 8.5dB at the test with TEST INPUT. (Analyzer photo)

There is no tunable component.
The resonant freq was adjusted by a parallel inductance (270nH) to the main inductor (15nH).

Keiko, Kiwamu

We noticed that we have used wrong code for MICH degree of freedom for both of the ELOG entries on this topic (cavity lengths tolerance search). It will be modified and posted soon.

 Keiko, Kiwamu

Length tolerance of the vertex part is about 5 mm.

Sorry for my procrastinating update on this topic. In my last post, I reported that the length tolerance of the vertex ifo would be 2mm, based on Kiwamu's code on CVS. Then we noticed that the MICH degrees of freedom was wrong in the code. I modified the code and ran again. You can find the modified codes on CVS (40m folder, analyzeDRMITolerance3f.m and DRMITolerance.m)

In this code, the arm lengths were kept to be ideal while some length offsets of random gaussian distribution were added on PRCL, SRCL and MICH lengths. The iteration was 1000 times for each sigma of the random gaussian distribution. The resulting sensing matrix is shown as histogram. Also, a histogram of the demodulation phase separation between MICH and SRCL is plotted by this code, as these two length degrees of freedom will be obtained by one channel separated by the demodulation phase. We check this separation because you want to make sure that the random length offsets does not make the separation of these two signals close.

The result is a bit different from the previous post, in the better way! The length tolerance is about 5 mm for the vertex ifo. Fig.1 shows the sensing matrix. Although signal levels are changed by the random offsets, only few orders of magnitude is changed in each degrees of freedom. Fig.2 shows that the signal separation between MICH and SRCL at  POP55 varies from  55 to 120 degrees, which may be OK. If you have 1cm sigma, it varies from 50 degrees to 150 degrees.

Fig. 1 Histgram of the sensing matrix including 3f channels, when sigma is 5mm. Please note that the x-axis is in long 10.

 Fig. 2 Histogram of the demodulation phase difference between MICH and SRCL, when sigma is 5 mm. To obtain the two signals independently, 90 is ideal. With the random offsets, the demodulation phase difference varies from 55 degrees to 120 degrees.

My next step is to run the similar code for LLO. 

Here are the results of the arm loss measurements, which I have done before the vent.

I ran the existing matlab script, called 'armLoss.m', to estimate the loss. The script resides in /scripts/LSC.

(Y arm)

 Round trip loss =  154.668624 +/- 11.343204 ppm

The figure above is a time series of the measurement.

In the lower plot the power in the ASDC_PD are plotted. The green dotted-curve is the power when the Y arm is unlocked.

The blue dotted-curve is the one when the Y arm is locked.

In the upper plot the estimated loss from each combination of locked/unlocked power are plotted.

(X arm)

Round trip loss = ????? 50 ppm ?????

The obtained time series looked wired because difference in the ASDC power when the arm was locked/unlocked were small.

This small difference results in such a small loss.

To see what was going on I will look at the trend data.

The measurement itself wasn't good.

I looked at the full 2 kHz data which was taken during the time when I was running the arm loss script on the X arm.

The plot below shows the raw data. The X arm was locked and unlocked sequentially several times.

The ASDC power didn't show a significant difference between the state where it is locked and unlocked.

I am not sure why, but It could be because of a misalginment or some kind of mode-mismatching, which can decrease the coupling efficiency of light going into the cavity.

(some notes)

The raw data were analyzed.

I split the ASDC data into two data, (1) low power state, when the cavity is locked (2) high power state, when the cavity is unlocked.

Then each state was averaged to estimate the averaged ASDC power in each case.

The number I obtained are :

    ASDC when X arm was locked = 54.77755 cnts

    ASDC when X arm was unlocked = 55.45830 cnts

Those numbers correspond to a round trip loss of  78.780778 ppm, which sounds too small for me.

 To check the demodulation boards for REFL33 and REFL165, a long cable from ETMY (SUS-ETMY-SDCOIL-EXT monitor) is pulled to the rack on Y side.

(1) A filter just after the RF input and (2) transfer function from the RF input to the demodulated signal will be checked for the two 3f demod boards to confirm that they are appropriate for 33 and 165 MHz.

There is a LP filter just after the RF input of an demodulation board (its schematic can be found as D990511-00-C on DCC). I have checked if the 3f freq, 33MHz, can pass  this filter. The filter TF from the RF input to RF monitor (the filter is between the input and monitor) on REFL33 demo-board was measured as shown in Fig. 1. At 33MHz, the magnitude is still flat and OK, but the phase is quite steep. I am going to consider if it is ok for the PDH method or not.

 Fig. 1 Transfer function from the RF input to RF monitor on the REFL33 demodulation board. At 33MHz, a very steep phase is applied on the input signal.

The phase delay due to the RF input filter on the demodulation board will not bother the resulting PDH signals.

I quickly calculated the below question (see the blue sentence in the quote below). I applied an arbitrary phase delay (theta) due to the filter I measured, on the detected RF signal by the photo detector. Then the filtered RF signal is multiplied by cos(omega_m) then filter the higher (2 omega_m) freqency as the usual mixing operation for the PDH signal. As a result, the I signal is delayed by cos(theta) and the Q signal is delayed by sin(theta). Therefore the resulting signals and its orthogonalitity is kept ok. From the sideband point of view, theta is applied on both upper and lower and seems to make the unbalance, however, as it is like a fixed phase offset on both SBs at the modulation frequency, the resulting signals is just multiplied by cos or sin theta for I and Q, respectively. It won't make any strange effect (it is difficult to explain by sentence not using equations!).

I lined up the Y Arm for locking and then centered the oplevs for ETMY and ITMY.

* The ITMY OL has still got the old style laser. Steve, pleaes swap this one for a HeNe. Also the optical layout seems strange: there are two copies of the laser beam going into the chamber (??). Also, the QPD transimpedance needs to be increase by a factor of ~10. We're only getting ~500 counts per quadrant. Its worth it for someone to re-examine the whole ITMY OL beam layout.

* The ETMY OL beam was coming out but clipping on the mount for the ETMY OL HeNe. This indicates a failure on our part to do the ETMY closeout alignment properly. In fact, I get the feeling from looking around that we overlooked aligning the OL and IPPOS/ANG beams this time. If we're unlucky this could cause us to vent again. I undid part of the laser mount and changed the height on the receiving mirror to get the beam back onto the QPD.

I noticed that there is significant green light now getting into some of the IR PDs; beacuse of this there are weird offsets in the TRY QPD and perhaps elsewhere. We had better purchase some filters to tape over the front of the sensitive IR sensors to prevent the couplling from the green laser.

* There is a beam on IPPOS, but its too big for the detector (this has always been the case). We need to put a 2" lens with a weak focusing power on this path so as to halve the beam size on the detector. Right now its clipping and misleading. There is also a 0.9V offset on the SUM signal. I'm not sure if this readout is working at all.

* I couldn't find any beam on IPANG at all. Not sure what's changed since Kiwamu saw it.

 The 2nd beam from this laser is for the SRM's OpLev, so that shouldn't be changed.

For better or worse, we didn't do anything to the ETM OpLevs, because they don't have any in-vac steering optics.  We did however go through and check on all the corner OpLevs.

A comment :

Since the LSC RFPD have a long cable of more than 6 m, which rotates a 33 MHz signal by more than 360 deg, so the delay has always existed in everywhere.

The circuit you measured is a part of the delay existing in the LSC system, but of course it's not a problem as you said.

In principle a delay changes only the demodulation phase. That's how we treat them.

RA: Actually, the issue is not the delay, but instead the dispersion. Is there a problem if we have too much dispersion from the RF filter?

Filters at the RF inputs of REFL33 and REFL165 demodulation boards were measured again. The filters will be totally fine for 33MHz and 165MHz.

Last time I forgot to calibrate the cable lengths, therefore the phase delay of the measurement included the cable lengths. This time the measurements were done for REFL33 and REFL165 demod board with calibration. As the cable lengths were calibrated, the shown plots (Fig.1 and Fig.2) do not include the phase delay dues to measurement cables. Please note that the x-axis is in linear. The phase delays of both boards seems to be not too steep (it will not affect anyway, as Kiwamu pointed out in his comment on the previous post). You can see that the two filters do not filter 33MHz and 165MHz component out.

Fig.1 A response of a filter which is placed just after the RF input of the demodulation board for REFL33. X-axis is shown in linear (~50MHz).

Fig.2 A response of a filter which is placed just after the RF input of the demodulation board for REFL165.

I also quickly checked the orthogonality of the demodulation board for REFL33 and REFL165 using function generators and oscilloscope. I checked the frequencies at 1,10,100,1K,10KHz of the demodulated signals. They are fine and ready for 3f signal extraction.