40m QIL Cryo_Lab CTN SUS_Lab TCS_Lab OMC_Lab CRIME_Lab FEA ENG_Labs OptContFac Mariner WBEEShop
  PSL, Page 38 of 52  Not logged in ELOG logo
ID Date Author Type Category Subject
  745   Fri Dec 2 20:03:50 2011 taraNotesFSSmodulation index for TTFSS

I got the wrong model on the EOM. It is actually model 4004, not 4003. Based on the specification, the modulation index is closer to what we expected (measured ~ 0.15, estimated ~0.09).

I checked the modulation index of the sideband.  It is unusual that we had measured the index of ~0.06.

      I checked the power going to EOM, to make sure that there was nothing wrong, and found out that the power dropped by ~8dBm when it passed through the RF summing box. I replaced the box with a spare one, and the power to the EOM remains almost the same, see figure below.


After replacing the box, I rechecked the modulation index again using two methods.

  •  Check the error signal, by scanning the laser and observe the error signal from mixer out. The pk-pk values of the carrier and the sideband's error signals are 16mV and 8mV respectively.

This give the ratio of Psiedband/Pcarrier to be 1/2 which means beta is ~ 1.

  •  Measure the transmitted power through the cavity between the carrier and the sideband. I measured both power by a power meter and V out from TRANS_PD. The ratio between Psideband/Pcarrier is still ~ 1.8 x10-3. This corresponds to beta ~ 0.1 (see the plot in the attachment). The result is higher than the previous result with the old summing box (0.06),  but still very small.

(The manual says that the modulation depth is more than 0.2 rad/Volt, see figure below, so if we have 160 mW coming in, with 50 ohm impedance => V = sqrt(RP) = sqrt(0.16*50) ~ 2.4V. So Beta should be ~ 0.2 * 2.4 =  0.48 rad. not 1.8 x10-3.)  This part is wrong, the modulation depth is  15 mrad/V see specification for 4004 below, the calculation is in the correction below. 



I suspected that the RF summing box might have a problem, may be the impedance match is not 50 ohm. We have not really checked every components in the spare box. 

     Next step, I'll remove the RF summing box, and connect the signal from LO to the EOM directly, and lock the cavity with FAST feedback only. If this gives us higher modulation index, it proofs that RF summing box does not have 50 impedance output to EOM and should be fixed.

==Correction as of Mon Dec 05 17:39:04 2011==

    I connect the power from LO directly to the EOM, and lock the beam using FAST feedback only. The ratio between sideband power and carrier power (Ps/Pc) is = 19.5uW/ 3.45 mW ~ 0.006 which corresponds to beta ~ 0.15


    The EOM we use is New Focus 4004 which has modulation depth = 15 mrad/V. The impedance is 22 pF (measured), and specified as 20-30 pF in the data sheet. For 35.5 MHz signal, the impedance is abs(1/ 2 pi 22 pF * 35.5MHz) ~ 200 Ohm. 

    Then for 22  dBm power (170mW), assuming all the power is transferred to the EOM, V =sqrt(PR) = sqrt( 0.17 Watt * 200 Ohm) ~ 6V -> Beta = 6*15mrad = 90 mRad. This is ~ a factor of 2 smaller than the measured modulation depth (0.15).

   Then let's calculate the input power we need for Beta = 0.6.  The voltage across the EOM must be 0.6/15mrad per volt = 40V. power = V2/R = 402/200 = 8 W -> 39 dBm .

 So to sum up,

  1. the modulation index increases from 0.1 to ~ 0.15when I remove the RF summing box and lock the laser with FAST feedback only.
  2. the modulation index is close to what we expect from the datasheet. To increase more modulation depth, we need more power from the LO.
Attachment 2: Snapshot_2011-12-03_01-17-33.png
Attachment 3: beta.png
Attachment 4: Snapshot_2011-12-05_19-21-10.png
  744   Thu Dec 1 22:00:53 2011 FrankSummaryRFAMEOM foam insulation (2nd generation) and temp stabilization

below some pictures of the second generation of EOM foam insulation for the room temp cavities. The main idea behind this version was to get a foam box with large enough, particle free beam holes which is following the EOM while being aligned. So this version's reference plane is the top of the EOM. The foam box consists of two different foams. One is the very stiff, old yellowish polyurethane foam used as the top plate. The main EOM insulation is the same white packing foam as used for the 40m version.


The tubing for the holes are parts from the Push-to-Connect Tube Fittings used for our air distribution for the table.


To electrically insulate the RF connection from the aluminum tape on the outside i've build a small insulated SMA extension.


For the heater i used the same 1" x 1" Kapton heater as used at the 40m. I have two temp sensors attached to the case of the EOM, one 1k Platinum RTD and one 100k NTC. The NTC has a much higher sensitivity. The RTD is currently used for temp calibration.


I've had a sample of one of the TeamWavelegth WTC3243  temp controllers which i used for a first try. They claim something like 2mK/24h stability but i don't know anything about their noise. My test board has an on-board reference which is currently used for the setpoint, but it also comes with an external analog input. Output current can be individually limited for positive and negative direction (when using a TEC). For use with a heater one of them is simply tuned to zero. The chip only has P and I but also comes with an "actual temp" monitor output (error-point monitor, but there is no out-of-loop sensor). Below a picture of the test setup.


  743   Thu Dec 1 19:35:04 2011 taraNotesEnvironmentSprinkler Installation

The sprinklers are installed in the lab, and I cleaned the lab afterward.


  742   Wed Nov 30 23:15:35 2011 taraDailyProgressNoiseBudgetISS

 ISS is on, I use RCAV_TRANS_PD to close the loop. Currently, we have only one way to actuate through the EAOM. We will use AOM for ISS soon, so that both intensity noise on each cavities can be controlled individually.


I use SR560 as a gain control for feedback to the EAOM, which is installed before the PMC.

The setup on SR560 is AC coupling, DC pass, with gain of 100.



fig1: RIN from RCAV is reduced by ~ a factor of 8 once the loop is closed (cyan -> blue) . However, there is not much improvement on ACAV's RIN, as the coherence between the two cavities are quite small.


The Noise suppression level is as expected. The plot below show the transfer function of the plant times the control gain in comparison with  RCAV_RIN(open loop) divided by RCAV_RIN(close loop). The pole around 33kHz in the plant TF is from the cavity pole.


fig2: Comparison between the expected OLG TF of the whole system( measured TF from plant times gain setup on SR650) and the result from the measurement (free running noise divided by suppressed noise). They agree well.

  741   Wed Nov 30 19:00:42 2011 taraNotesNoiseBudgetvertical seismic coupling to beat noise

  From 40m page, COMSOL FEA gives us the coupling (from vertical seismic to frequency noise ) to be 53 [kHz / (m/s2)].


Fig1: TF measurement( from PSL:735). This time the estimation from FEA is plotted as well.


  •      From ACAV/seismic (Green), the measurement is ~ 1order of magnitude higher than the FEA result. This might come from the fact that the support is not exactly at the right place
  •      From RCAV/seismic (red), the TF tells us nothing. The coherence is close to zero, and the SNR is very low (~10-6 - 10-5). This requires 1010 average to bring SNR to 1. Below is the plot for Signal (frequency noise due to seismic) and noise (NPRO free running noise. And SNR (Signal/Noise)


     fig 2: Signal and Noise in TF measurement between RCAV noise and seismic.

        The NPRO free running noise is from psl:617. Seismic noise is measured by an accelerometer (PSL:716)and converted to frequency noise using coupling from FEA result (53kHz / (m/s^2))

  ==what's next==

  1.  Since the SNR for RCAV noise/seismic is very small, we might have to shake the table and remeasure it.
  2.  Use a better seismometer. The current accelerometer might not be sensitive enough.
  3. Check the coupling from FEA, check what is the coupling as a function of supporting position.
Attachment 2: Seismic_resp.fig
  740   Tue Nov 29 18:40:09 2011 taraDailyProgressNoiseBudgetISS

 I measured RIN from RCAV_TRANS_PD, ACAV_TRANS_PD, and PMC_TRANS_PD, and coherence between the three signals. It looks like that RIN from ACAV and RCAV are mostly from pointing (from DC upto a few hundred hertz), rather than intensity fluctuation.



There are 5PDs in total, 2 for ACAV_TRANS, 2 for RCAV_TRANS (see PSL:738), and 1 for PMC.  The gain on ACAV and RCAV PDs are set to 20dB so that the outputs were not saturated. Dark noise from every PD(not shown) is approximately 50 nV/rtHz flat, which is well below the signal at all frequency.

The plot below shows:

  • RIN from RCAV_TRANS (behind RCAV) (BLUE) as measured from RCAV_PD1 and RCAV_PD2. They are similar (as it should be).
  • Likewise, RIN from ACAV_TRANS (RED)from ACAV_PD1 and ACAV_PD2 are the same. This is just to check that the PDs work fine.
  • RIN behind PMC in GREEN
  • Coherence between RCAV_TRANS_PD and ACAV_TRANS_PD (magenta)
  • Coherence between ACAV_TRANS_PD and PMC_TRANS_PD  (Yellow)
  • Coherence between RCAV_TRANS_PD and PMC_TRANS_PD  (Cyan)



      From the shape of the measured RIN, Frank suggested that we might be suffering from pointing effect. Since we pick the beam after the cavities  any pointing instability will cause the transmitted power to fluctuate, and this will show up on RIN measurement.

      Coherence between PMC and RCAV (Cyan) at ~ 200Hz and above is almost 1, this should be caused by real intensity noise. However at low frequency the coherence is close to zero. Any power fluctuation observed at the transmitted beam from RCAV is not really caused by the fluctuation at PMC. Pointing effect can be one explanation.

       I have to mention that the AOM on RCAV path is not being used. So the beam just travel through the AOM. But we still use AOM on ACAV path to frequency shift the beam. The first order beam which is frequency shift can be suffered from pointing effect if the alignment of the reflected beam does not trace back the incoming beam. Pointing from AOM might explain the lower coherence between ACAV and PMC (Yellow).

==To do next==

    To confirm that it is mostly pointing effect, we'll use the pickup beams for RFAM for RIN measurement. Since these beams are picked up before entering the cavities, the intensity noise due to pointing effect should be reduced. We expect to see better coherence between PMC and RIN infront of the cavities.   Unless we fix the pointing, ISS won't improve any RIN measured from TRANS_PD.

Attachment 2: RIN.fig
  739   Tue Nov 29 01:50:40 2011 taraDailyProgressopticoptics for RFAM

I made some minor adjustment to the optics layout so that the reflected beam at the PBS before the cavity can be used to measure RFAM. Now RCAV's beam can be picked up for RFAM measurement.

    The PBS just before RCAV was moved Eastward a bit so that the reflected beams from both PBSs are not blocked. I removed mirrors with soft mounts and use only rigid 1" posts only. 

    I used a spare 35.5MHz RFPD for the pickup beam from RCAV path (in red). The power cable for RFPD was made and checked. It works properly. There is a spare new focus 1811 RFPD, but the connector is broken, the pins are bent. I'll try to fix this and use it for ACAV's RFAM pickup.


     The AC signal from RFPD will be demodulated with 35.5 MHz signal which is split from the LO signal for ACAV PDH's lock. I have not adjusted the phase by trying different cable lengths yet. This will be done later.

     There is one thing I'm a bit concerned with. The RF signal from the RFPD has DC level ~ 120 mV, I'm not sure if it's unusual or not. I'll check with another RFPD.

  738   Tue Nov 29 00:40:01 2011 taraDailyProgressNoiseBudgetISS

Photodiodes for ISS are all setup and are ready to be used.


All four PDs are Thorlabs PDA10CS InGaAs amplified Detectors. Since InGaAs is not very sensitive to room light, I do not put room light filters over the PDs.

For each beam path (RCAV/ACAV) two PDs are required, one for feedback signal, one for out of loop measurement. The one for feedback signal can't be used for taking the noise level because it is in loop and the signal is suppressed by the control gain, thus does not reflect the real intensity noise of the laser.

Next step: verify the noise level of the RFPD, measure RIN from the PDs.

  737   Wed Nov 23 01:22:48 2011 taraDailyProgressNoiseBudgetISS

The previous entry was wrong. I did not use another PD to pick up the suppressed noise. So I fixed the setup and add another PD for picking up the signal.


 ==designing the control for the loop==


The above diagram is a block diagram for a simple ISS. We can measure dP (free running noise). The suppressed noise (dPs) is picked up by another PD when the loop is closed. To design the TF of the controller, we have to know what are the TF of each compnents, so the whole TF can be decided.

I measured the TF of actuator and PD ([A] and [D]) in the diagram. The setup was hooked up similar to the drawing, but SR560 was set to AC coupling, A-B, flat response for injecting the excitation. The PD's gain can be adjusted, the measurement below was taken when the gain was 20dB.


The gain on the PD can still be adjusted, I have to check this again.


The next step is to decide how much gain do we want. We know what is the RIN (measured from last entry) for the current setup, but we have to estimate how much it couples into frequency noise, then the gain level can be decide.

  736   Mon Nov 21 21:38:36 2011 taraDailyProgressNoiseBudgetISS

 ISS was re-installed to the setup, with the EAOM as an actuator. I used SR560 as a temporary servo. A better servo card will be design and installed in the electronic crate later.

==motivation & background==

       Relative intensity noise (RIN) of the laser, and other mechanisms that cause power fluctuation, can couple into the frequency noise of the laser locked to a Fabry-Perot cavity via random absorption. Thus, we want to take precaution and reduce it by applying an intensity servo to the setup. Our current intensity actuator set is consisted of an EAOM followed by a polarizing beam splitter. An EAOM changes the polarization of the beam, then the power transmitted through the PBS is changed accordingly. 

==current setup==

I used a Thorlab PD to dectect the transmitted beam from RCAV and measured RIN. Then I used the same signal to feedback to SR560 and to the EAOM. The setting on SR560 is AC coupled with bandpass with cutoff frequency at 10Hz and 3kHz, gain 1k.




==comments & future plan==

I'll check what other groups have with the intensity noise and project it in our noise budget. For now the only group that report the effect of intensity noise I known is Alnis' group. I'll also add the result from our measurement as well.

Attachment 1: RIN.png
  735   Thu Nov 17 23:30:19 2011 taraNotesNoiseBudgetvertical seismic coupling to beat noise

TFs between 1)vertical seismic and beat, 2) v seismic and ACAV feedback, and 3) v_seismic and TTFSS feedback are measured. The contribution from vertical seismic to beat signal is mostly from ACAV.


I used frequency response measurement (FFT1/FFT2) to measure frequency response [dB], phase [dB], and coherence between.

  • vertical seismic and beat, -> D/A
  • Actuator point for ACAV loop -> C/A
  • Fast Actuator point, RCAV loop ->B/A  (I used a T connector to pull the signal from feedback to NPRO)

The calibration for beat signal[D] is 7kHz/V, for ACAV actuator point [C] is 7kHz/V, for RCAV fast actuator [B] is 3.07 MHz/V, for accelerometer [A] is 1023mV/g (psl:716)



 1)calibrated signal, in the unit of Frequency noise in beat/ acceleration  [Hz/ (m/s^2) ]


fig1: calibrated frequency response, with 250 vector avg.


  The plots show that the seismic contribution in the beat is mostly coupled through ACAV. RCAV is not sensitive to seismic at all, since the coherence is almost zero. We have to open the chamber and fix the ACAV's support.

Attachment 3: freq_resp.fig
  734   Wed Nov 16 19:11:38 2011 taraNotesNoiseBudgetseimic coupling to beat noise

Seismic noise is currently a limiting source for our setup, here is a list for what we have to do for figuring out the coupling from seismic to beat. Once we know exactly what's going on we can fix it correctly.

  1. TF measurement between seismic and feedback to NPRO (RCAV)
  2. TF measurement between seismic and feedback to VCO    (ACAV)
  3. TF measurement between seismic and beat             (psl:)

[more details soon]

  733   Tue Nov 15 18:53:25 2011 taraNotesNoiseBudgetupdated number for everything

A list of useful numbers for noise budget calculation.


==Reference Cavity==

  • Cavity Finesse:  10^4    PSL:424
  • Free Spectral Range:  737 MHz <=> cavity length = 8" = 0.2 m
  • Cavity Pole = 38.7 kHz.
  • FWHM = 2xcavity Pole = FSR/Finesse = 73.7 kHz
  • Spotsize on mirror = 291 um


==PMC==  PSL:370

  •  Finesse: 188
  • FSR     : 714MHz
  • FWHM: 3.8MHz
  • P-polariztion


==35.5MHz RFPD==

  • RFPD noise level:


  • Fast actuator calibration factor: 3.07 MHz/V psl:182
  • Slow actuator


  • Tuning range vs frequency psl:22


  • Marconi(IFR2023b) calibration factor = 70e3  Hz/V for 160MHz carrier/ 100kHz input range (tuning range)
  •  Marconi(IFR2023b) calibration factor = 7e3 Hz/V   for 160MHz carrier/  10kHz input range

==Seismic Stacks==pslL668

  • translation beamline: f0 = 6.96 Hz  Q = 21.53
  • Vertical motion          f0 =  Q= 
  • Horizontal transverse, f0 = 6.35Hz, Q = 25.85


==ACAV servo==

  732   Tue Nov 15 18:38:20 2011 taraNotesFSSmodulation index for TTFSS


By measuring the power of the transmitted carrier and sideband, we can compute the modulation index of the laser.

P_carrier = P0 x J0(Beta)^2, and

P_sideband = P0 x J1(Beta)^2 . (P0 is the input power, P0 ~ P_carrier + 2 P_sideband), J0,J1 are Bessel functions of the first kind.

By plotting P_sideband/P_carrier vs Beta ,we can see where our modulation index is.

==setup & measurement==

  1. The RF_adj = 9.5V
  2. Phase adj = 0V
  3. Use old RF summing box

 The Transmitted power of carrier, and sideband are 8.3 mW, and 5.7 uW respectively. This gives Ps/Pc ~ 1/1000, or Beta ~ 0.06. This is surprisingly low. I'll double check it again.

  731   Mon Nov 14 21:51:07 2011 taraNotesFSSFSS debugging, OLG TF measurement

I remeasured OLG TF of TTFSS, the bandwidth can be increased upto ~300 kHz for 12 mW power input. It seems that the servo gain is certainly will be the limiting source for the current setup. We will need a higher gain to suppress laser frequency noise down to Brownian coating thermal level.




    I locked the laser to refcav, measured noise at error point, and converted it to absolute frequency by the slope of error signal. Then I compared it with the beat signal. Then I checked the OLG TF of the TTFSS and its  UGF. I decided to check it again, since the beat signal now became better than before.


 1) beat and noise at error point



fig1: Beat, with comparison from noise @ error point. Note that harmonic lines appear again. I think because I use external power supply on the TTFSS. I'll use the power supply from the electronic shelf later.


The bump around 1kHz can be reduced further by increasing more gain, but not much. However, it was almost unstable.

    2) OLG TF: Details about how to measure TTFSS OLG TF can be found here(PSL:)


fig2: OLG TF of TTFSS, at three different input power, the plots from lower input power (1.6 mW and 6.6 mW) are offset to compare with the highest input power (12mW). The UGFs for 1.6mW, 6.6mW, and 12mW input power are ~1kHz, 2kHz, and 3kHz respectively.

From previous measurements  ( psl:594, psl:654), the UGF was quite low (less than 200kHz), and the TF looks terribly noisy.  Now with much better TF, I'll use this result and compare it with LISO model.


The UGF of 300 kHz is quite high compared to what we had before (I have not yet measured the actual value of UGF for power input of 12 mW, I just estimated by eye. The real value might be a bit less.) However, from the current setup, it is might be too hard to increase more power to RCAV, because ACAV will not have enough power.  An RFPD with higher gain might be needed, since we will need at least a gain 2 order of magnitude more than the current one so that the suppressed noise can be down to coating noise at 1kHz.

Attachment 2: beat.fig
Attachment 4: OLGTF.fig
  730   Thu Nov 10 16:37:19 2011 taraDailyProgressNoiseBudgetRFPD noise in PLL

 I measured noise level from PD for beat measurement. It will not be the limiting source at frequency below 1kHz.

 ==back ground and motivation==

see psl:728


 (i) dark noise :from the diagram in psl:728, I just disconnected the feedback signal from actuator to Marconi, blocked the beam and measured from Vfeedback to actuator. The signal is similar to the signal from Verror (after the mixer) multiplied by gain from SR560.

(ii) shot noise: A white light source was setup on the PD such that the DC level from the PD was similar to what the regular beam would produce (0.47V). Then repeat the measurement as in (i)



     The signal at Vfeedback was converted to frequency noise by dividing by SR560 TF, and frequency discriminator TF (gain 20, Low pass @ 10kHz and 0.45/f V/Hz respectively) so that the Voltage noise from RFPD is converted to absolute frequency noise.

      The frequency noise and shot noise from PD go up with frequency and crosses the estimated coating noise around 1kHz. The current PD should be sufficient for the setup for  measuring coating thermal noise at frequency below 1kHz.


  729   Thu Nov 10 00:05:45 2011 taraNotesSeismicPressure in N2 bottle


 One of the leg has a leak a gain. It's the one on SouthWest corner of the table. When I tried to floated the table, I can hear the air leaking from the leg. That leg is a slave one, which has no pressure control, so I'm not sure if this happened because I tried to adjust the pressure on NorthWest leg to balance the table or not. The table on NW leg was too high and the leg height was pushed up against the pin and not "floated".

  728   Wed Nov 9 20:31:40 2011 taraDailyProgressNoiseBudgetRFPD noise in PLL

I measured PLL's open loop gain(OLG) transfer function. This will be used to determined the noise contribution from beat RFPD.  For the current setup, PLL has UGF ~20kHz.


I have not really checked how beat PD's noise will show up in the beat signal. I want to make sure that it will not be the limiting source. After discussing with Koji, I did some calculation and planned how to measure the noise level.


 fig1: some calculation for PLL's block diagram. [H] -> frequency discriminator, [F]-> Servo (SR650), [A]->Actuator(Marconi)


==setup== [add fig]


 During the measurement, I reduced the gain of SR560 to 5, so the signal at low frequency was not suppressed too much. And the UGF was reduced to ~ 9kHz. (For regular setup, Input range = 10kHz, SR560 gain = 5e2, LowPass at 10kHz).

During close loop operation, I measured:

  1. Open Loop Gain of the TF [H]x[A]x[F] as shown in the setup, and
  2. Actuator x Frequency Discriminator[H]x[A], the response to SR785 was taken after the mixer (and the low pass filter)




The OLG TF can be fitted with a pole at zero, and a pole at 10kHz( from SR560).

The pole at zero comes from the combination of Marconi and Frequency discriminator [H]x[A] (green plot). I think Marconi's TF [A] has flat response, at least up to a few kHz, since all the calibration we have used so far is flat, and it matched up with LO noise up to a few kHz. For examples, psl:506, psl:459. 

(I'm not sure if Marconi really has flat response or not, but let's assume it's true for now, and the pole comes from the frequency discriminator [H] only. I think [H] will be flat response if we convert it to V/rad by multiplying by f)

Green plot is [H]x[A] which is 20*log10 (3.16e3/f) or the actual transfer function is ->[H]x[A] = 3.16e3/f [V/V]

Since [A] = 7e3 [Hz/V] and flat, [H] is then-> 3.16e3/f / 7e3 = 0.45/f [V/Hz]

Once we get [H] we use it to calculate frequency noise contribution from Verr or Vfb when the feedback to Actuator is not connected (loop is not closed)(cf fig 1)

Attachment 3: TF.fig
  727   Wed Nov 9 02:10:58 2011 taraNotesFSSFSS debugging, harmonic line problem fixed

 From PSL:725, it seems that beat signal at high frequency is limited by  FSS servo gain. So I took a quick check by switching from old FSS to TTFSS. There is some improvement at high frequency. I'll try to float the table to see if I can reduce the bumps at frequency above 100 Hz or not.



  • The LO phase noise from PLL (10kHz input range)does not match the red plot (TTFSS, 10kHz input range) I'll have to check the noise budget again.
  • When I used 1kHz input range (green plot), I used calibration = 0.7 kHz/V. This was not measured, I just scaled it from other measured calibrations form other input ranges ( 100kHz -> 70kHz/V, 10kHz -> 7kHz/V). The result seems ok, the features between blue and green curves line up nicely.
  726   Tue Nov 8 18:51:38 2011 taraDailyProgressopticoptics for RFAM

I added mirrors to pick up stray beams just before the cavities. These beams will be used for monitoring RFAM.

    I arranged the optics so that stray beams at the beam splitters (just in front of the cavities) could be used. The power of the beam is ~ 9 uW, but it can be increased by changing the polarization of the input beam later.

     Two photodiodes are needed, I haven't checked yet if I still have some spare PDs left.

     Then the signal from PD will be demodulated with 35.5 MHz signal (modulation frequency). The cable length + PD position will be adjusted so that the phase is the same as the PDH signal.




  725   Tue Nov 8 03:15:10 2011 taraDailyProgressFSSFSS debugging, harmonic line problem fixed



From psl:722, the beat noise at high frequency was getting a bit worse from before. I brought it back to previous level. It turned out that misalignment at AOM, was the problem.

    I had to re align the beam in ACAV path a lot, some mirrors might be touched when I adjusted the cables around that area. After that the beat level came back close to what it was.



Fig1: Beat, compared to error point from RCAV and ACAV loop.


  • Noise from RCAV's error point (at high f) is very close to the current beat level and LO phase noise, but it seems that seismic noise also present around high frequency. We may have to identify what causes those peaks around 200 Hz and above first.
  • Noise from ACAV's error point is just slightly above the noise floor of SR785(not shown), so this might give only the upper limit of ACAV's error noise.
Attachment 2: beat_2011_11_07.fig
  724   Sun Nov 6 22:54:57 2011 taraNotesNoiseBudgetRFAM/ ISS

A list of things I plan + have to do this week

 1) RFAM 

  • Pick up a stray beam somewhere for RFAM measurement when the cavity is locked. Demodulate the signal to get the time series of RFAM level.
  • Add RFAM to noise budge.
  • Design a case to cover the EOM for temperature control.(Take a few pictures of the EOM)

2) ISS

  • Add intensity servo. Use AOM to actuate on the power of both beams to ACAV and RCAV (might start with the exist EAOM first).
  • Use a simple servo system for a start, the final design will be a servo card installed on the electronic shelf.
  • Add the effect of RIN induced noise to the noise budget
  723   Thu Nov 3 20:36:42 2011 taraNotesSeismicPressure in N2 bottle

I checked the pressure of N2 bottle used for floating the table. I floated the table again today for measuring beat. The pressure of the bottle does not decrease much. It is still ~ 2000 psi. I closed the valve on the bottle after I finished the measurement.

  722   Thu Nov 3 19:51:07 2011 taraDailyProgressFSSFSS debugging, harmonic line problem fixed

  I identified that harmonic lines in ACAV loop come from UPDH box. When I used SR560 to lock ACAV loop, there was no harmonic. Now the lines in beat signal are suppressed considerably. 

      From previous entry psl:720: I eliminated the harmonic lines in RCAV loop (FSS). I tried to bring the harmonic lines back in ofrder to see what caused them, but I could not. So I'll just have to remember to check the RF phase adjustment if the lines come back again. 

     After the lines in RCAV loop were gone, the beat signal still contained harmonic lines, so I check ACAV loop. The procedure is similar to what I've done for checking FSS loop. The lines were gone when I replaced the UPDH servo with SR560. Then I measured the beat from that setup.



Left: Current beat signal (in blue) as ACAV loop was locked by SR560 instead of UPDH. Right: Close up at high frequency, line harmonics are greatly suppressed.



[add fig


The beat signal becomes slightly worse at high frequency. This might be a result from:

  • Not enough gain in ACAV loop (since I changed the servo), or
  • Scattering noise/back reflection. ( The visibility of ACAV reduced to~50% from 90%)

    I think  because I plugged and unplugged the cable connected to the EOM all the time, the EOM might be mis-aligned, I might also touch the mirror for double passing AOM on ACAV path as well. Right now the visibilty for ACAV become ~ 50% (from 93%), the visibility for RCAV decreased as well, but I brought it back to ~95% again.  I could not bring the visibility for ACAV back to the same level. I have to check what's wrong with the setup. This misalignment might be the reason for worse beat signal.

Attachment 3: beat_2011_11_03.fig
  721   Thu Nov 3 01:55:03 2011 FrankNotesSeismicring down measurement for rubber spring

The total mass of the load including the accelerometer was 397g. Will recalculate the spring constant an damping of the RTV later.


  720   Wed Nov 2 20:05:15 2011 taraDailyProgressFSSFSS debugging, harmonic line problem fixed

 I got rid off the harmonic lines in FSS loop.  I'm still trying to understand what actually caused the harmonic lines.


 Fig1: PSD from error point of FSS loop. Blue, with harminic lines, Green, when it is fixed.



       I started the setup similar to the setup in the latest test. psl:711 . However, I used the spare RFsumming box ( the pomona box in the picture)


Fig2: RF summing boxes on the table, the spare one is sitting on the stage above the table for easy access.

Then I continued to investigate by adding the following components to the loop:

  1. cable for EOM feedback, from the electronic rack to RF summing box
  2. 35.5MHz oscillator card from the rack instead of a Marconi to drive EOM and mix signal for the mixer.
  3. Original cables in the setup, (a) from RFPD to mixer, (b)from oscillator to EOM,
  4. FSS without then with PC feedback

     After all the above tests I still found no harmonic lines, so I switched back to the original RF summing box, and found all the lines.  I though it might be coupling from the cables close to the box, so I move the summing box up where I placed the spare box, but the harmonics still presented.


      Then I adjusted the phase for the LO signal to the mixer so that the error signal was optimized. After I optimized it, the harmonics line were gone. I think the phase adjustment is what I have to look into. I'll try to understand that more later.

     Next step will be trying to lock ACAV with SR560, make sure there is no harmonic line, then check beat measurement to see any improvement at high frequency.

  719   Tue Nov 1 16:11:48 2011 taraNotesopticV-block for faraday isolator

I made a drawing for faraday isolator's base. I'll submit the drawing tomorrow.

Attachment 1: isolator_vblock.PDF
  718   Fri Oct 28 14:04:22 2011 frank, taraNotesNoiseBudgetNoise budget for unfloated table

 We think it might be useful to see the seismic contribution to the beat signal when the table is unfloated (so that the seismic contribution can show up clearly). Here is the beat and noise budget when the table is not floated.


For vertical seismic noise, I used measured seismic noise on the unfloated table, converted to acceleration noise, then applied the calculated TF (psl:697)of the stack and the result from comsol (acceleration noise to length noise). That is

[Measure Seismic converted to Acceleration] x [TF of the stack] x [COMSOL]  and converted to absolute frequency noise using df/f = dL/L.

[m/s^2] x [x/x] x [ m/(m/s^2)] x [Hz/m],

Note: This is the contribution from a single cavity, no common mode rejection is taken into account.



For East West seismic (beamline), I obtained that by using the measured TF (psl:716)between beat and signal from accelerometer, applied calibration for acceleration, multiplied by measure seismic on the unfloated table (converted to acceleration) x calibration for beat.

[Measured TF beat/acceleration] x [seismic converted to acceleration] x [calibration for acceleration] x [calibration for beat (7kHz/V)]

[V/V] x [ m/s^2] x [V/(m/s^2)] x [Hz/V]

Note: The mismatch between E-W seismic and the measured beat might come from the measured TF. We drove the table with the PZT and this may cause some non-linear response at the resonant frequency which shows up in the plot.

Attachment 2: beat_27.fig
  717   Thu Oct 27 20:45:58 2011 frank, taraNotesSeismicring down measurement for rubber spring

We did ring down measurement of the rubber cone used for supporting the seismic isolation stacks. The resonant frequency and Q of the spring will be used for better TF of the stack later. The measured values are f =46.7 Hz, Q = 14.67. (mass will be added later)

[details about the measurement will be added later]

These are the results from the data Frank gave me.

There is a problem in the result. The frequency of the ring down response can vary from 42 Hz to 47 Hz. So when I fit it I cannot really get the "best" frequency fit from the whole train of data.

From figure (1)

  • At point A) The data and fit seem to be well matched.
  • At B) The phase of the fit curve advances the measurement ( the fit frequency is too large)
  • At C) The phase of the fit curve lags the measurement (The fit frequency is too small)
  • At D) The frequency totally changes causing the fit and the measurement to be almost out of phase


fig1: showing fit with the data. Non-linearity behavior of the spring can be seen clearly.


There are 5 sets of measurement. I notice that the nonlinear behavior was very small in the first data set which has its initial amplitude smaller than 0.4 V


fig2 Fit from data TEK00000. The initial amplitude is less than 0.4 V.


For other data sets, where their initial amplitude exceed 0.4, it does not matter where I analyze the data, fitted frequency still varies a lot. I take the data set #4, and  fit the data at high amplitude and low amplitude (when it is less than .15V). The frequency still varies. See fig3 below.


fig3: fit from data TEK00003. Even when I use data at smaller amplitude, frequency still varies from the fit, similar to what happens in fig(1).


The problem is, the fitted frequency from fig(2) is ~ 46 Hz, while the fitted frequency (at low amplitude) from fig(3) is ~ 43 Hz. The difference is too large. So we decided to sample the frequency by averaging the frequency over 3 adjacent cycles of the ring down data, ignoring the first few peaks. Then histogram the samples.


Most of the counts are around 46.5 Hz, which is contributed mostly from the first data set where the variance is very small. I think it is ok to pick the value from that data set

f =46.7Hz, Q = 14.67.

  716   Wed Oct 26 18:05:49 2011 frank, taraNotesSeismicModeling of the vertical isolation of the stack

Today we floated the table, measured beat, measured TF between beat and horizontal, and added horizontal seismic to the noise budget.

  • Replaced 2 legs, floated table: monitor pressure, 2000psi @ Wed Oct 26 18:06:41 2011
  • Measure TF between beat and accelerometer signal
  • measure beat signal, when table is floated, optimized optics.
  • add Horizontal motion to noise budget
  • ring down measurements from rubber spring, see psl:717

details coming soon.



fig1: accelerometer was mounted on the table to measure East-West motion (along beam line)


fig2: details of the accelerometer


fig3: TF between Beat signal and signal from acceleromter (East-West motion). The data is not calibrated yet. Sensitivity for beat = 7 kHz/V.


fig4: Current beat noise: floated table. Seismic from horizontal motion is calculated by applying the measured TF to the measured acceleration on the floated table.


fig5: Current beat noise: floated table. Seismic from horizontal motion is calculated by applying the measured TF to the measured seismic noise (converted to acceleration) on the floated table.

Attachment 4: beat.png
  715   Tue Oct 25 23:21:55 2011 Frank, TaraDailyProgressBEATTF from hor. acceleration to beat signal

as we still don't have a good model for the horizontal isolation of the stack we will try to measure the TF from horiz. accel. of the table to the beat using the PZT actuator between steel frame and table. With the table resting on the legs we only get a nice TF around the resonance of the stack at ~6.5Hz, the rest is too noisy. We will fix the broken legs to float the table tomorrow and then try again.

  714   Tue Oct 25 23:13:57 2011 Tara, FrankDailyProgressSeismicre-measured the spring constant for the RTV springs

as we only did a quick-and-dirty measurement of the spring constant last time we re-measured it a couple of times. The resonance frequency varies by a few Hz from measurement to measurement and it also depends on over how many periods (and where, beginning or more to the end) of the ring down we measure. So we will analyze some ring down measurements using the same setup (Al-block with accelerometer resting on one RTV spring sitting on the optical table) and take the average for a new calculation.

  713   Tue Oct 25 22:03:32 2011 frank, taraNotesSeismicModeling of the vertical isolation of the stack

From Koji's code. I added vertical seismic noise from a single cavity, no common mode rejection. There are some modification from the code. The spring constant is 45 e3. M1 = 12.3, m2 =5.44.

The seismic noise in this projection is the data taken last week, see psl:696.


The coupling between acceleration noise to length noise is1e-9 * L  [m/(m/s^2)]. It is different from the value from FEA because our supporting positions are not exactly the same as the optimum point. It might be off by a few mm. The peak around 6.7 Hz is from Horizontal motion which has not been incorporated in the plot yet.

The noise budget code and auxillary files & functions will be committed on svn.


  712   Mon Oct 24 21:20:26 2011 frank, taraDailyProgressFSSrf summing box

 We check modulation depth of the PDH signal to see if we have enough modulation depth or not. The result will be analyzed later.


Based on the design of the RF summing box we got from Peter ,


the TF from 35.5 MHz input to EOM does not have a resonant frequency at 35.5 MHz. Rather, the resonance is ~ 10MHz. We are not sure if we need to have the resonance at 35.5 Mhz or not. So we decide to check if the modulation index is already large enough for the current design, and compare it with that of the modified one. If the modulation depth is already large enough, we can keep it as it is. Basically we just need a high pass for 35.5MHz input to EOM.

By changing the inductance L from 1uH to 80 nH, the peak can be moved to ~ 35.5 MHz.

==Setup & Measurement==

 We want to check the slope of the error signal when the RF power changes, so there's a small change in the setup, see fig1.


The input power is 2mW.

Then the slope of the error signal was record, the pk-pk of the sideband and carriers were also record between the input power of 10dBm and 13dBm to the EOM.

Then the RF summing box was modified (changing L), and the measurement was repeated.


coil       |  input pwr (dBm) |  Carrier slope ( MHz/V) |   pk-pk value (mV) (carrier/side band)          

 80nH   |     10                   |     1.43                          ||   28/18

            |             13           |      1.11                          ||  38.8/22.6

1 uH    |       10                 |      0.91                          ||   49.2/27.6

            |             13          |       0.74                         ||    68.4/37.2


From the result, the best setup is 1uH coil with maximum side band power which provides the best slope ( maximum gain at frequency discriminator). Note that 1uH coil is the original value for the schematic, so modification to 80nH does not help.

-> Modulation depth

  711   Fri Oct 21 01:36:52 2011 frank, taraDailyProgressFSSFSS debugging/ checking

We locked RCAV with a very simple FSS servo to debug what components in the FSS are not working correctly. For now, we found out that RF summing box does not work as expected and reduce the detection gain.



We decided to check the RCAV fss loop seriously in order to understand why the performance is not as expected. So we used a very simple setup an SR560, a marconi, with FAST feedback only to lock the laser and measured the in loop noise of the laser. Then add more components in the setup and see how it changes the performance.



                 set 1: The laser is locked with SR560, VCO: Marconi 2023A.  All cables used were extra cables not from the original setup. The in loop noise was measured with gain setup = 1e3 and 2e3. 2e3 gain level was barely stable, but we added it for comparison.

We checked the slope of error signal, and optimized it by changing the length of cable A and B, to change the phase between incoming signals to the mixer. Note that the signal sent to EOM were not 50 Ohm terminated. It turned out that with 50 Ohm, we got smaller gain.  The calibration for this setup at error point is 1.2 MHz/V. [show result]

               -set 2: RF summing box is added to the setup




  1. The setup for SR560 was changed.
  2. The gain (slope of error signal) was reduced, once we added the RF summing box to the system. The slope was ... ,
  3. I could increase more gain (up to 5000 without losing lock)which further reduce the in loop noise at low frequency, but a peak at high frequency ( ~70kHz) appear. At this point the time series of in loop noise was oscillating. And 180 Hz harmonic also showed up. This is a good example of non linearity in the system.
  4. The length of A and B cables were adjusted again to obtain the best error signal. 



Plot1: uncalibrated in loop noise from the above setups. Notice the absence of higher harmonics line when the system was not oscillate (red, blue and cyan). I did not save the data up to 100kHz to see the peak over there since our bandwidth of interest is only around 1 Hz - 1kHz.


Plot2: calibrated signal to absolute frequency. I omitted the result from Seup2 gain 5000 (magenta), because the system oscillates. Results from other setups are kept in the same color. As we can see, adding the RF summing box setup 2(red) causes the overall frequency noise to be higher.

It's true that the gain is different by a factor of 2, which might be the reason for lower gain. But we have differnt slope from error signal, and the system oscillates at higher gain, so the RF summing box definitely changes the loop performance.


==RF summing box==

   Our goal is to fix the RF summing box so that the gain is not reduced and the in loop noise level is maintained. This is the schematic of RF summing box we got from Peter King.

[add fig]

We also checked the spare one we have in the lab.  [add fig]. Components on 35.5 Mhz input path are similar to the schematic. We have not checked the tank circuit on the feed back path yet, but that's not necessary for the current test.

Attachment 4: inloop_volt.fig
Attachment 6: calibrate_inloopnoise.fig
  710   Thu Oct 20 11:43:36 2011 taraNotesNoiseBudgetcalculation from kenji's paper added to the nb

 Result from Kenji's paper (in brown) is added to the noise budget.



[Numata etal., PRL 93, 250602 (2004)]


Attachment 3: nb_2011_10_20.fig
  709   Wed Oct 19 19:32:20 2011 Tara, FrankDailyProgressSeismicleg of optical table replaced - table now floating- beat

 Today we measured frequency response between beat signal and seismic noise, and coherence when the table was not floated. The data are calibrated based on the information from 40m:5196 (with gain 1 for out setup).(The plot is not calibrated yet)


     We wanted to learn more about how seismic couples to beat measurement for the current setup. So measuring the frequency response between beat signal and seismic should tell us about the TF between the table to the cavities. For example, we can put a noise budget due to seismic when the table is floated by measuring the seismic on the floated table, then multiply by the TF we measure, plot it on the noise budget.


==setup & calibration==

    Details about the seismometer we used can be found here psl:696.  The table was not floated. The direction and positions of the seismometer (pink circles) are shown below in the map. All the results in this entry come from position 1 only. We checked that the signals of seismometer did not change much among each position.

Signal from seismometer was sent to SR785 chA. Signal from beat was sent to chB. We used Frequency response measurement (B/A) with vector average. The TF between beat and seismic noise from 3 directions were recorded along with coherence between them.

The beat signal has calibration = 70 kHz/Volt. Seismometer has calibration of 800 [Volt/velocity] or [Volt *sec]/ [meter].

The result (recorded in dB [V/V]) can be converted to [Hz / (m/s) ] by 10 dB/20 x 70 k [Hz/Volt] x800 [Volt / (m/s)]





The top panel shows the magnitude of the transfer function. Phase is shown in the middle panel. The bottom panel presents the coherence between seismic and beat for each direction. We are quite surprised that there is no strong peak around 6-7 Hz which is the resonant peak of the stack motion. We certainly see the peak around 16Hz which probably comes from bounce mode of the stack.

Attachment 2: TF_beat_seismic.fig
  708   Mon Oct 17 20:26:57 2011 taraDailyProgressFSSGround Loop, RFPD

Today I verified that fss is not gain limited. I also tried to fix the ground loop problem in fss by replacing the power supply, but it did not work.


 ==FSS gain limit==

     From psl:706 I mentioned that I did not check the in loop noise level vs gain. So I checked it again.The in-loop noise (MIXER OUT signal when the RCAV is locked) does not change at all when the gain reaches 25 or more. There's a change as the gain was increased from 20 to 25.  This means the noise is probably from the detection point. Modifying/changing RFPD will be the next thing on the list.

 ==Ground Loop==

   From psl:700, I noted that the ground loop might come from the power supply which had ground output same to protective ground. I found a power supply which has float ground and replaced it with the current one. Then I measured MIXER OUT(in loop noise) when the cavity was locked. 

 The results:

  1.  In loop noise look similar from either power supplies I use.  Odd harmonics (3f, 5f, ...) are still present.
  2. There are no even harmonics(2f,4f,...) in rcav loop. I found it to be from acav loop.


==beat PD noise level==

 I just realized that I have not checked and thoroughly understood the noise level of the beat PD and how it shows up in the beat noise. This will be the next thing to do as well.

  707   Mon Oct 17 12:01:26 2011 taraNotesSeismicN2 bottle ordered

I called central warehouse to order another N2 bottle.  It should be here this afternoon. Then we can find which leg has a leak.

  706   Fri Oct 14 14:40:51 2011 Tara, FrankDailyProgressSeismicleakage in the legs, RFPD noise test

I think another leg, or both of the old legs have leakage. The nitrogen tank is now empty.  So the latest beat signal (in blue) here was measured when the table was not floated, but the optics are optimized.

I also tried to lock the cavity with both servos (old fss and TTFSS), The beat signals look similar (the result was not saved, only through inspection by eyes), so either noise from detection point (RCAV RFPD) or noise from ACAV loop might be the limiting source at high frequency.


 == beat signal after optic optimization==


           Fig1: Current beat signal (blue), the table was not floated.


== RFPD==

As a quick test, we checked the RFPD, by measuring the in loop noise at error point (MIXER OUT) with the current RCAV's RFPD. Then compare the results with the RFPD Raphael modified (see here). Although we don't know the performance of the current RFPD, we expected them to be somewhat different.

I used old FSS for this test, input power is 2mW, Common gain = 30 (max), Fast gain =15.6.  Higher Fast gain causes the system to be unstable.



Result 1: The noise level at error point: The noise level is ~ 100 nV/rtHz. at high frequency. Not particularly good. Both RFPDs give very same noise level.


Result2: Noise at error point calibrated to absolute frequency noise, compared with beat signal.

As both RFPDs give the same noise level at the error point, this should mean that the gain of both RFPDs are the same, and the calibration factor to absolute frequency will be similar in both cases. I measured the error signal slopes to be 3.6 e5 [Hz/V] and 2.9 e5 [Hz/V] for current and modified RFPD respectively. I decided to use measured calibrations from each case in the plot so that the level of noise due to possible error bar from the measurement can be seen. The level of error point noise is quite close to the beat level (a factor of ~1.4 - 1.6)

This means that we might be limited by RFPD noise at high frequency.  It might be something else. More investigation will be done. (I did not see if the noise level change with the gain or not at the current setup, I just cranked the gain up to what ever it was before it started to oscillate. If the noise still goes up/down with the gain, it's not RFPD noise)


Attachment 1: beat_2011_10_13.fig
Attachment 5: errnoise.fig
  705   Thu Oct 13 02:36:58 2011 Tara, FrankSummarySeismicsecond leg replaced

After optimized everything, I repeated the measurement that was done in this entry (noise at detection point). There is some improvement, the result is shown below.


==What have been done==

By "optimize everything", I meant:

  1. The Faraday Isolator was installed back in the setup, and optimized for maximum isolation. [add pic]
  2. Beam splitter and quarter wave plate sets (for double passed AOM and cavities) were optimized for minimum back reflection.
  3. EOM with half wave plate were adjusted for minimum RFAM (reduced by ~ 20 dB)
  4. Beam alignment to the cavities: visibility are up to 93% for both cavities now.

 Measurement Recap:  We want to check the noise from scattering noise or RFAM at the detection point, so we measure the noise at error point when the beam reflected off the cavity, or a mirror in front of the chamber (cf entry:700). We also want to see the dependent on power input, so we chose 10mW, 5mW and 1mW input power.

    Below are measurements from error point (Mixer out) which are calibrated to absolute frequency noise through the slope of error signal from each setup (power input of 10mW, 5mW and 1mW)


==Comments for the results==

  1.    The noise level when the beam reflected off the cavity goes down when we float the table and optimize everything ( red, green, blue are lower than pale pink (result from 2011_10_07).
  2.    The noise level when the beam reflected off the cavity do not change much with input power ( red, green, blue are about the same)
  3.    At 1mW the noise at high frequency (above 50Hz)  raises up for both cases (reflected off cavity / mirror) I'm not sure what happens. The calibration seems to be ok since the noise level at low frequency matches the results from 5mW and 10mW setup.
  4.    Seismic stack's resonant frequency at 6.7 Hz shows up more clearly after RFAM/ back reflection are minimized.


To Do: We will check the beat signal tomorrow. There should be improvement, since optics are optimized and the table is floated.

Attachment 2: plot_2011_10_12.fig
  704   Thu Oct 13 00:24:35 2011 FrankDailyProgressRefCavnext iteration of refcav stack

we got the stainless compression springs and i cleaned them together with the all the other parts for the radiation shield/heater assembly (except the heater wire). The only thing i'm not sure about is how to attach the temp sensors (ad590, metal/ceramic package) to the copper shields. I would like to solder them on it, probably using indium. I don't like to glue them as i don't want to bake the stuff (we don't have a large enough oven so i would have to wait for a spot at the 40m). I could just put it into our spare chamber and pump on it for a while. The other parts are small enough to fit into the baking chamber.

Something i'm not sure about is: What happens to the springs (stainless steel) if baking them at let's say 120C for a few days? Do they change their hardness? Does anyone know?

I also wanna try to cut holes into the RTV springs to make them softer by removing material. We checked the cork drill's at the 40m. Most of them are broken. Looks like someone used them to open bottles or for cutting metal
So i probably buy a set (~$12 for a set of 6), but could also make one for the diameter i want to try (it's only a thin-walled tube with sharp end, so very easy do make if the machine shop has the right tube).

Before i bake the stuff i will try different versions with the spare stack on the hepa bench (which has not the right size of top plate, but we can add some load)

  703   Wed Oct 12 00:53:40 2011 Tara, FrankSummarySeismicsecond leg replaced

As the table was floated, we measured the noise from error point again.


     We tried to determine if the noise bump we saw were from the window, so we place an extra window in front of a mirror [add fig] and compare to the noise when there was no window. The results are not different that much.

     From a quick look, by adjusting the input power, from 1mW to 10mW. The shape of the noise from error point changes substantially. This could be come from RFAM or scattering.  I'll measure the noise vs input power after I optimize everything first. RFAM, beam splitter, back reflection have to be optimized.



  702   Mon Oct 10 21:30:45 2011 Tara, FrankSummarySeismicsecond leg replaced

we have replaced the second leg which was leaking. It could be that the legs are simply to old and the rubber got brittle. As far as i know the table has never been operated floating as it had the suspended reference cavity on it since the beginning. We operate at around 85 PSI, maximum operational pressure is specified at 100PSI so that should be OK. The second leg started leaking after one day of floating operation. We disassembled one today to have a close look but we can't really tell where the leak is. We will check with some pressure to see what's broken within the next few days. Let's see what the other legs do in the near future. We still have plenty of "spares", as Aidan bought two new sets (taller ones) for the old tables in the TCS lab. So we have 8 short ones which are currently not used (and seem to be newer than the ones we currently have). And they still work as passive legs.

  701   Mon Oct 10 11:17:46 2011 FrankNotesPurchasesnitrogen bottle replaced

we got the new nitrogen bottle this morning. We could check the price for nitrogen and compressed air bottles to see if it's worth getting a regulator for compressed air bottles in the future. But i think we should stick to the nitrogen. Once we fixed the leaking leg we shouldn't need much air anymore...

  700   Sat Oct 8 02:09:56 2011 frank, taraDailyProgressFSSNoise Hunting: Ground Loop, RFPD

Today we did noise hunting stuff. We focused on ground loop and noise from RFPD. The noise from ground loop is probably come from the power supply of the RFPDs, and noise from RFPD looks very suspiciously similar to beat. More investigation is needed to be done.

  ==Ground Loop Problem==

     We saw a lot of odd harmonics from 60 Hz peak (180Hz, 300Hz, and so on). This indicates that we have a ground loop somewhere. We find that the table has the same ground from the cables (which come from the electronics shelf). We tried unplugging all the cables that connected to items on the table ( PDs, RFPDs, EOM), but it did not fix the ground loop. We have not tried unplugging the laser's cable yet. However, we think we know where the ground loop comes from. The power supply for the RFPD are three pins +/- 15 V with ground. Its ground is the same as power supply's ground which directly connects to the RFPD.  To fix this, we have to use a power supply from the electronics shelf, and we haven't tried to fix this yet.  This will be tested later.


==Noise at High Frequency==

We switched back to old FSS to compare the noise level. [add fig, details]


==Noise from RFPD==

   We checked the signal from error point of RCAV FSS loop. It did not changed with power input of the laser (from 1mW to 10mW), so we think it might come from the detection point, RFPD. We check the noise at error point of the FSS loop (old FSS) when

  1.   the beam incident on RFPD was blocked, so only electronics noise is observed, and
  2. the beam reflected from the cavity (off resonance) and fall on the RFPD, so we know what is the effect from the incident beam on RFPD (RFAM, scattering etc) and
  3.  a mirror is used to reflect the beam back to the RFPD in front of the cavity, so we know what is the effect from the incident beam without contribution from the cavity.


 Fig1: setup to check noise from detection point.


We measured the slope of the error signal (1.1e5 Hz/V) which is used for converting the noise at error point to absolute frequency noise.

The result is shown in comparison with beat measurements (red and magenta) below:


Note: I have only coating noise in the noise plot so that the graph does not look to busy. The whole plot can be found in the matlab fig flie attached below.

Comment about the result:

  1.      The result from case (2) (beam reflected off the cavity) has a shape very similar to that of beat measurement. It is only about a factor of 10 away from the beat. If we assume that the calibration we got from error point could be off by a factor of 3, and the noise in beat measurement includes the similar effect from ACAV's path which might have larger calibration. It is very likely that light falling on the RFPD is the origin of the bump at 100 Hz. 
  2.      When we used a mirror to reflected the beam of in front of the chamber (case 3), the noise went down considerably. This means that the scattering bump come from the window of the chamber or the cavity itself. From inspection, we think it may be the window, because it is not very clean. A new AR window might be needed. 
  3.     The electric noise at the detection point (case 1) is already higher than coating thermal noise level. RFPD has to be modified later.



To do: fine adjustment on beam splitter and waveplates, to reduce an

Attachment 3: noise_hunt.fig
  699   Fri Oct 7 14:07:26 2011 FrankHowToVacuumsmall baking chamber

i've build a small baking chamber which we can use for small parts like posts, sensor and heater assemblies, wires, screws etc. It's a short (10" long) 6" CF reducer Tee with two heaters and a k-type thermocouple on it. We still don't have a RGA, but if we clean all parts carefully (made of materials we know that they are not a problem in general) and vacuum bake them for a few days we reduce the risk of contaminating our cavities by a lot. Current wait time for parts submitted for baking is ~2month. Would be great to find a slightly larger chamber to fit the larger parts in there as well (i don't want to use the (now) spare refcav chamber).


  698   Thu Oct 6 15:30:14 2011 taraNotesPurchasesNitrogen bottle

We can buy a tank of Nitrogen/ pressurized air  from Caltech warehouse, the extension is x4891. They have delivery service as well.

  697   Thu Oct 6 03:37:44 2011 KojiNotesSeismicModeling of the vertical isolation of the stack

The vertical motion of the stack was modeled.

  • In the current configuration, the two vertical resonances are to be located at 13.6Hz and 48.8Hz with Q of 63 and 18
    (Aha! You can see the suspicious peaks in the latest beat noise spectrum.)
  • If the upper stage is suspended by four Stainless Steel spring (Frank suggested McMaster Carr 9435K91), the resonances move to 3.3Hz and 34.7Hz.
    The upper stage will be almost completely decoupled and will have very high Q that is determined by the loss of the SS springs.
  • Currently there is no vibration isolation at 10Hz, and a factor of 200 at 100Hz. This will be improved to 10 and 10^-4, respectively.

Model parameters:

Current configuration

Mass of the upper stage, m1: 14.27 kg --- Frank's estimation
Mass of the lower stage, m2: 5.44 kg --- Frank's estimation
# of elastmers at the upper stage, n_spring1: 8
# of elastmers at the upper stage, n_spring2: 8
Spring constant of the elastmer, k1&k2: (0.396)*(2*pi*43)^2 --- RTV615 block, 43Hz resonance with 0.396kg mass
Damping factor of the elastmer, gamma1 & gamma2: 0.396*(43*2*pi)/20 --- Q=20 assumed based on M. V. Plissi, et al, Rev. Sci. Instrum. 69, 3055 (1998)

Modified configuration

# of elastmers at the upper stage, n_spring1: 4
Spring constant of the elastmer, k1: 1529N/m --- McMaster Carr 9435K91 (8.75 lbs./inch)
Damping factor of the elastmer, gamma1: 0 (no loss)

The attached zip-file contains a Mathematica file with full derivation/calculation and Matlab files with TF calculations.

Attachment 1: vertical_isolation.pdf
vertical_isolation.pdf vertical_isolation.pdf
Attachment 2: vertical_isolation.zip
  696   Thu Oct 6 00:47:28 2011 Tara, FrankDailyProgressSeismicleg of optical table replaced - table now floating- beat

We measured seismic noise (on three directions) on the table when the table was floated and not floated. Seismic is substantially reduced at frequency 10 Hz and above.

Note: the data is calibrated to velocity/rtHz. ( I checked the 40m entry, I used this calibration here. The data was corrected for 50Hz pole as well.

(after calibrating, the result look similar to what Jan did last year, see here psl:435)


==guralp cmg-40T setup==

We used Guralp CMG-40T 3-axis seismometer (borrowed from 40m with Jenne's help) to measure the noise on the table. The setup on the seismometer is "Broadband velocity". Signals from "Low velocity" channels are used to acquire the data.

Voltage supply is 12V, the output ground of the box is fake ground and it should not be connected to the ground of the oscilloscope/ spectrum analyzer. I used float ground on the spectrum analyzer, so it should be ok.

The voltage output for each channel is +/- 10V.

The position was tuned by adjusting the legs of the seismometer, so that the position readout for each channel is close to zero (we got less than +/-0.3V), the manual says it should be below 3.5V, so we are ok.

==seismic improvement==

3 directions and the corresponding resonant frequencies are

  1. Vertical    (2.5 Hz)
  2. North-South: horizontal transverse motion of the cavity (normal to cavity's beam line)  (1.8 Hz)
  3.  East-West:  beam line direction       (1.7 Hz)


The seismometer has flat response up to 50Hz, the data has been corrected for the transfer function of the seismometer.

***The noise is getting worse at low frequency below 3 Hz is typical [link to Newport], and the signal above 50Hz are mostly noise, so both signals from floated and unfloated table are similar. (2011/10/06)


Transfer function from the floor to floated table. [Newport]


===beat improvement===

Then we checked the beat signal. We just made a quick check to see if the peaks between 10-100Hz would change or not, so we did not try to optimize the loop or anything. The result (red) improves, those peaks are decreased significantly, but it is bad that noise at frequency above 100 Hz goes up a little bit. This might be from the un optimized servo. So, the next step will be checking the servo, noise at high frequency.



Attachment 1: seismic.png
Attachment 2: seismic.fig
Attachment 4: beat_2011_10_05.fig
Attachment 7: calibrated_seismic.fig
ELOG V3.1.3-