Today, we completed the power boards and tested to make sure they work. We had a problem with the LM2991 negative voltage regulator; its tab at the back is connected to the input voltage and should therefore not touch our panel, because it is in this way grounded. We will fix that problem by placing spacors below the power boards.

We also built the connections for our DAC and ADC controllers and designed our DAC & ADC panels so that we can order a company to drill holes for our wires.

To this point, we have to finish the circuits for some of our Hall-effect sensors boards (our missing components arrived) and test all of our circuits to make sure they work as desired. After that, we will move to the second stage of this summer research, taking measurements, debugging problems, improving computer processing, and ultimately successfully isolating the levitated plate up to very low frequencies (starting from 5Hz and hopefully going down to 0.01Hz)

I returned the Gooch & Housego R26035-2-1.55-LTD AOM (SN 216939) from DOPO table

The low frequency noise looks pretty good now. The funny shape is most likely a thermal transient due to having not enough insulation. You need to droop some Kleenex over the circuit to stop the thermal air currents and then put a second box over the first box. Then its probably best to sit outside of the room when taking the measurement to reduce the human noise.

Norna and Rich: I am sorry for taking so long to get you this measurement. I plan to do the noise measurements on the standalone LEDs this week.

The following table gives the current through each OSEM's LED (measured using the voltage drop across the 238-ohm resistor in series), as well as the measured photocurrent (the DC output of the amplifier divided by its transimpedance gain of 100,000 V/A), and the ratio of the two. The plot from the previous post is reproduced below for analysis--I realized that I did still have this plot saved in units of V/rHz. In some cases (e.g. #3, #4), the noise level seems to be correlated to the photocurrent, but not all of them follow this pattern. The issue of #1 being significantly quieter than the other set remains, as well.

 Tonight, I calibrated the AOSEM's response in [A/m]. I used a sophisticated rig consisting of:

1. One of those anodized Faraday isolator mounts to hold the OSEM

2. A translation stage with a screw gauge to jam something into it (gracefully)

3. Some DC power supplies and multimeters

4. My simple transimpedance amplifier sketched in the previous post (I did not bother upgrading the readout circuit for this measurement since I was just taking a relatively rough DC measurement)

---------------

I used a 9/64 hex key to simulate the shadowmaking magnet. A picture of the setup is attached below.

Some pertinent info:

- The current through the LED was maintained at 35 mA throughout the measurement. The measured voltage across it was 1.62 V, giving Z_LED = 46.3 ohm.

- The op amp supply voltage was +/- 10 V, and the PD bias was +10 V.

- The output voltage of the amplifier with the PD fully lit was 3.04 V (measured before and after the test). Note that this voltage increases slightly as the key is inserted due to reflections. 

The second attachment is a plot of the photocurrent versus the position of the key (the x axis is shifted such that the key is roughly centered at x = 0, and x < 0 corresponds to the key being further inside). The response of the OSEM in the linear region is roughly 0.05 A/m.

  I was able to take some better measurements last night. I took data in two bands: 0-100 Hz, 0-1.6 kHz, each with 800 lines. This gives us a decent idea of what's going on at low and high frequency. Attached are four plots, two from each band. All measurements were taken with a box over both the OSEM and the readout circuit and the lights out.

The first two are low- and high- frequency comparisons of the noise in the full (bright) configuration as measured with no BPF and AC coupling vs with the BPF and DC coupling. There appears to be no difference apart from the expected effect above the pole at 1 kHz.

The next two are plots of the noise in various components and the full scheme calibrated into equivalent displacement noise. Everything is below ~10e-10 m/rt(Hz) with the exception of line peaks, and again it would appear that we are limited by our measurement equipment.

Some notes:

- The "dark" noise seems to be coincident with the "amp" noise with the exception of some extra pickup that increases at high frequency (seems to be line-related).

- The "LED" noise is coincident with the "supply" noise up until its 8-Hz corner frequency, after which it falls off as expected until it hits an apparent floor around 100 Hz.

- The "bright" noise seems to be coincident with the "supply" noise, while the "dark" and "amp" are much lower. This could be because the supply noise only shows up when there is an appreciable voltage at the output of the amp.

Have to think about this for a bit, but the next logical step is to turn the measurement setup into something solid (i.e. soldering, enclosure, etc.).

On Friday, Rana and I discovered that my transimpedance amp was oscillating like whoa at about 100 kHz. A little research showed this to be due to the input capacitance of the AD743 (~20 pF). To fix this, I put a 20-pF cap in parallel with the 100k feedback resistor, and that seemed to do the trick.

The relevant circuitry is shown in attachment 1. +/- 12 V DC was provided by voltage regulators (7912, 78M12). The voltage across the LED was measured to be V_LED = 1.61 V, and the current through it was I_LED = 31.6 mA, giving Z_LED = 50.9 ohm. The voltage out of the amp with a fully lit PD was V_out = -2.83 V, giving a photocurrent of I_ph = 28.3 uA. 

I was concerned about noise that might be imposed by the bandpass filter, so I compared spectra I took with and without it (that is, AC coupled, no BPF, and DC coupled, with BPF). This comparison is shown in attachment 2. There appears to be no difference apart from the aliasing effects at low frequency.

After this, I took the real measurement, extending the range to 800 Hz, averaging 100x and with a linewidth of 1 Hz (I realize now that I should probably have done this with a smaller linewidth, so that I could see below ~1 Hz. I will repeat the measurement this week with better low-frequency resolution). The result can be seen in attachment 3, calibrated to displacement noise in m/rt(Hz) using the measured 0.05-A/m response of the OSEM in the linear region. The four lines are:

- Bright: noise in the OSEM with a fully lit PD

- Dark: noise in the OSEM with the LED off

- Amp: noise in the transimpedance amp with the input terminated

- V_LED: noise in the LED voltage

The first three spectra were taken at the output of the amplifier and calibrated back to meters using the transimpedance gain and OSEM response in A/m. The last was taken across the LED, and calibrated into meters using the values given in paragraph 2. All measurements were taken with the OSEM under a box and with the lights out.

It appears that we are still limited by our setup. The "Dark" noise is coincident with the amplifier noise, while the "Bright" noise is coincident with the LED noise. That said, it is fairly comforting that all this noise is at the level of around 10^-10 m or less, as we can probably expect the true noise of the OSEM to be lower than this. We will know this for sure once we have a truly quiet setup (starting with ultra-low-noise voltage references).

 I tracked down some more AD743s at Wilson House 2.0 today (thanks to Rich). I was then able to simultaneously use 743s to both drive the LED and amplify the readout. Below is an 800-line DC - 25 Hz noise spectrum of

- The voltage across the LED (DC level: V_LED = 1.59 V) -- AC coupled

- The output of the amp with a fully lit PD (DC level: V_out,full = -2.95 V) -- DC coupled, through bandpass filter

- The output of the amp with the LED out -- DC coupled, through bandpass filter

- The output of the amp with an open input -- DC coupled, through bandpass filter

all calibrated to equivalent displacement noise. For the LED plot, this was done by using the measured current ratio between the LED and the PD when fully lit along with the measured OSEM response of 0.05 A/m. The other three were converted using this response along with the transimpedance gain of the amp (100,000 V/A). For all measurements, the OSEM was covered by a box, and the circuit was draped with a cloth and put under a box within another box (to reduce air currents).

The funny low-frequency junk from the previous driver spectrum is gone--thanks to the isolation from air currents--but the line seems a bit higher overall (trying to figure out why). There also still seems to be the funny effect of noise added by the LED above the level of the voltage noise across it, and I think it's somewhat strange that the "Dark" noise is lower than the amp noise with no input. We can probably still do better..

 It's dusty in here...

--------------------------------

I was recently commissioned to do some noise measurements on the new  AOSEMS. I set up a humble experiment in the LIGO e-lab to do some preliminary measurements:

I made a simple current-to-voltage converter out of an OP27E (using a 100-kohm feedback resistor) to use as the transimpedance amplifier for the readout. This results in a transimpedance of 0.1 V / uA. A simple schematic of the important elements is attached below.

DC power was provided without regulators directly from the laboratory DC supply in the lab. The value of 1.7 V across the LED was set such that the current through it was ~35 mA.

Rana and I took a few important PSDs (one of the DC supply, one of the OP27E with no supplied current, and two with the setup fully connected--one each with and without the PD covered), all from 250 mHz - 200 Hz, AC coupled. Using a sophisticated estimation method (called, by some, the "pick two points and approximate with a power law" method for lack of something fittingly elegant), we obtained a rough estimate of these spectral densities in order to compare them.

These were all converted into equivalent PD current noise. For all but the "supply" noise, this was done trivially by dividing by the transimpedance of the OP27E. For "supply", LED voltage noise had to be converted to PD current noise in the following way:

Z_LED = 1.7 V / 35 mA ~ 50 ohm

equivalent PD current noise = (I_PD / I_LED) * (measured supply voltage noise / 50 ohm)

where the PD-LED current ratio was found empirically to be (I_PD / I_LED) ~ 1 / 1000 by measuring the voltage out of the amp with full brightness (i.e. I_LED = 35 mA, no obstruction) and dividing by the transimpedance (see 2nd figure).

The third figure below is a plot of these spectral densities in common units. Somewhat expectedly, the noise of the "dark" configuration seems limited by the supply noise. However, the "bright" line seems to be dominated by something else. I'm not sure I see how it could be anything but the LED itself, but it is worthwhile to repeat this "test" with a better setup.

On the to-do list:

1. Voltage regulator/reference

Rana thinks that the AD587LN is a good choice of reference given its performance on some LISA tests. I am in contact with AD, and there is no longer a 'LN' package, but I am trying to get samples of the currently manufactured one that is most similar (AD587KNZ).

In the meantime, I am going to find some simple regulators downstairs or at the 40m.

2. Bandpass filter

I was advised that it is a good idea to build your own high/bandpass filter instead of relying on the spectrum analyzer's AC coupling function. I will be doing just this.

3. Switch to a better op amp

        Like the AD743

4. Calibration

I need to find a good way to hold the OSEM in place while I stick something in there with a micron drive without it being unreliably shaky.

We have LT1021-7 at the 40m, next to the Alberto's desk. This is the VREF for 7V.

EDIT: I have calibrated the y axis of the plot to meters

Last night, I got around to testing some of the other AOSEM samples, to see how the noise varied between them. What I found was rather strange: the noise in all the new ones (#s 2-6) was about the same, but they were all quite a bit noisier than the previous one I have been testing (#1). The only difference between them, as far as I can tell, is that the first specimen has a coil wound around it already, while the others just have a rubber band. Also, the newer ones all have an impedance of ~ 44-45 ohms, while I measured 47 ohms for the first (though, among the new ones, the slight variation in Z seems to have no correlation with the small differences in noise level). For those wondering, YES, I did remeasure the noise in the 1st one; I am not using old data.

Either my meddling with the old one has somehow made it quieter or something is amiss.

[Federico, Gabriele]

Unfortunately the measurement we performed last night didn't work well (microphone were off).

So we started again a set of measurements this afternoon with the blade loaded with about 16 kg as before, and peak to peak motion of about 500 microns.

The script stopper a couple of hours ago. I restarted it at about 4pm. The blade sagging trend seems to be stopped.

[Federico, Gabriele]

We left the system taking data for the whole weekend, and yesterday night Xiaoyue switched off the drive. Unfortunately, we discovered that the microphones moved during the weekend, at some time around Friday night one of the two fall off the blade. It turned out that petroleoum jelly is not a good option for us to attach the microphones to the blades. So, in conclusion, no good data over the weekend.

We then tested that using wax instead of jelly still gave us good acoustic coupling of the microphone to the blade and a solid mechanical connection. We tested two microphone side by side: one attached witth petroleoum jelly and another with wax. When touching the blade or hitting gently the table, we could see exactly the same signal in both sensors.

We then modified a Thorlabs angled bracket to host the blade: in this way we can launch it at about 60 degrees. When loaded with a 11 kg mass, the blade curves so that the terminal part is roughly horizontal. We installed again the shado sensor and the coil to drive the blade. The two microphone are attached close to the blade base, using wax. The excitation for the drive is generated by a DAC controlled with the KR2 model.

At about 7pm we started a python script that will continuosly turn on and off a sinusoidal excitation at 100 mHz, with intervals of one hour. The amplitude is such to have 200 um of displacement at the shadow sensor.

Friday September 18th from 19:00 LT: frequency 0.4 Hz, drive gain 1.0

Saturday September 19th from 10:00 LT: frequency 0.4 Hz, drive gain 0.7

Saturday September 19th from 21:25 LT: frequency 0.4 Hz, drive gain 0.3

Sunday September 20th from 8:45 LT: frequency 0.4, drive gain 0.15

Sunday September 20th from 21:42 LT: gain 0

[Gabriele, Federico]

At about 6pm LT we started a new run for the acoustic emission measurments with the high carbon steel blade. Load is about 13 kg, peak to peak amplitude of the motion is 200 microns.

After analysis of the full night of data, still no excess signal above the sensor noise was detected.

[Federico, Gabriele]

As discussed during the crackle meeting, we want to rule out the possibility that what we see is due to high frequency noise generated by the DAC or coil driver.

So we installed a passive low pass filter (corner frequency at about 30 Hz). We started a new run, excitation is a sinusoid at 0.4 Hz as before. Amplitude is 2 V, giving us a peak to peak motion of about 300 microns.

At 21:14 LT we started the drive on/off script, to collect one-hour-long stretches of data with drive alternating on and off.

Our trial to collect good data during the night for the blade acoustic emission faled: an error in the script I wrote resulted in periods of drive on and off lasting only 60 seconds, so too short for any analysis, since we collected only 20 periods in total. 

Looking at the MIC1 and MIC2 LF signals (not rectified) and acquired at 65kHz, it's clear that there isn't much there except noise. The following plot shows a histogram of 30 minutes of data: the behavior is very close to gaussian noise, as expected since the microphones aren't that sensitive below 30 kHz.

[Gabriele, Federico]

In past elogs we observed that when we were driving the blade, the histograms of the microphone demodulated signals were larger. As suggested during the crackle meeting, we suspected this effect to be a systematic due to noise introduced by the DAC or the coil driver.

So we installed a low pass filter (30 Hz corner frequency) right before the coil. Widening of the distribution was gone. 

Yesterday we removed the low pass filter and carried out another short set of measurements: we didn't get the histogram widening back. So the disappearance of the previously observed excess noise was not clearly due to the installation of the low pass filter.

We observed, without the filter, that when driving the microphone spectra show two additional lines at 17320 Hz and 13840 Hz (most likely aliased down from higher frequencies). Those lines are there only when we drive and their amplitude are modulated by the 0.4 Hz drive. 

Federico then modified the coil driver to have the low pass filter before the amplification. We ran all night and we can't see either lines of increase of histogram width.

Here is a summary of observations in different configurations:

1. Thursday night: no low pass filter, free blade, drive on and off comparison. Noise distribution was larger when driven, lines appear when driven and they are modulated at 0.4 Hz
2. Thursday afternoon: no low pass filter, blade blocked, drive on and off comparison. Noise distribution was the same regardless of driving state, but lines were there when driving, modulated as before
3. Monday night: low pass filter after coil driver, free blade, drive on and off comparison. Noise didn't change and lines weren't there
4. Yesterday afternoon: no low pass filter, free blade, drive on and off comparison. Noise didn't chanche, but lines appeared as before when driving
5. Yesterday night: low pass filter at coil driver input, free blade, drive on and off comparison: Noise didn't change, lines didn't appear

In conclusions, we believe that the appearance of the lines is uncorrelated with the increase of the histogram width. We can remove those lines with both configurations of the low pass filters. 

There is one additional thing we did in between the weekend and yesterday test: we removed the microphones for the face to face test, and attached them back. They are in the same position on the blade, as precisely as we could do. However, cables might be in a different positions, so we can imagine that they are stressed in a slighly different way. It's not impossible that we are seeing some effect due to the small bending of the cables (mechanical or electrical noise)

I almost completed an acoustic enclosure (48" x 24" x 12") for the laser board of crackle 2. It is built using Auralex 2" thick mineral fiber insulator , covered with white plastic panels. The wool parts are glued togetether with Loctite PL300, and the plastic panels are glued to the wool with the same glue. I finally sealed all the open parts with silicone rubber.

The parts are still drying in the electronics shop. Once cured, I'll clean everything up, cut some holes for cable feedthrough and move everything to the crakle lab.

The new acoustic enclosure is in place:

Acoustic noise as measured by the microphone is reduced by a significant amount, as visible in the following plot (blue no enclosure, red with enclosure)

The enclosure also improves the low frequency beam jitter (due to air currents)

[Saikanth, Xiaoyue, Gabriele]

Now we can easily lock the Michelson interferometer, so we started some noise hunting.

Bell jar

First we compared the displacement noise with and without the bell jar to close the vacuum chamber. Here are some times:

NO BELL JAR
Michelson locked - all damping off - not whitened signals
      Thu Jul  9 14:17:06 PDT 2015
      Thu Jul  9 14:18:35 PDT 2015
Switched to whitened (the signals are saturating from time to time)
     Thu Jul  9 14:19:12 PDT 2015
     Thu Jul  9 14:20:42 PDT 2015

BELL JAR IN POSITION
Michelson locked - all damping off - not whitened signals
      Thu Jul  9 14:35:46 PDT 2015
      Thu Jul  9 14:38:08 PDT 2015
Switched to whitened (no more saturated!):
      Thu Jul  9 14:39:40 PDT 2015
      Thu Jul  9 14:42:07 PDT 2015

Adding the bell jar provided a large reduction of noise at high frequency. See the first figure. Note that the calibration of these traces is good only above 100 Hz, where the locking loop has no more any effect. 

Seismic noise

We then clamped down the accelerometer to the base of the support structure, into the chamber. Again, we locked the interferometer. There is not very high coherence with seismic noise anywhere.

Acoustic noise

We then performed some acoustic noise injections using the loudspeakers in the lab and white noise from http://simplynoise.com/, with different volumes. Basically every structure above 100 Hz get larger, see the second figure. This isn't conclusive, but we can guess that most of the noise in the structures above 100 Hz come from acoustic excitations. 

Here are the times:

Michelson locked, all damping off, whitened signals: 
      Thu Jul  9 15:07:33 PDT 2015
      Thu Jul  9 15:09:34 PDT 2015
(blue trace)

White noise on (http://simplynoise.com/ max volume - speaker volume 47)
      Thu Jul  9 15:14:13 PDT 2015
      Thu Jul  9 15:16:04 PDT 2015
(green trace)

Back to non whitened signals (volume 55)
     Thu Jul  9 15:18:00 PDT 2015
     Thu Jul  9 15:19:24 PDT 2015
(brown trace)

Back to non whitened signals (volume 65)
     Thu Jul  9 15:21:55 PDT 2015
     Thu Jul  9 15:23:44 PDT 2015
(red trace)

Closing down the large ports

Finally, we closed the three 6"3/4 ports with three aluminum plates, as well as we could. The back one is used to get the cables out of the chamber, so we added a piece of carboard between the aluminum plate and the port. See the pictures:

This action had a large impact on the noise, expecially closing the back one, which is the one closer to the noisy cymac rack. The last plot shows the reduction in the displacement noise. This is properly calibrated, using a measured loop suppression transfer function.

As a reference, the measured open loop TF and suppression TF are shown in the last figure.

 Updates on active damping!

First, I performed a Voltage to Displacement calibration on the two shadow sensors, using a micrometer stage and a detector card to bring the sensor to half of its max voltage, and then measuring the voltage vs. micrometer displacement. The behavior was linear to within a few percent, so I feel confident in the conversion figure (different for each sensor, both both are close to 4V/mm)

This allowed me to turn the previously measured noise spectrum (incorrectly labeled PSD...) into displacement noise. Here's the plot:

Next, I adjusted some of the differentiator circuit values, to optimize for the relevant frequencies. Using liso, I calculated theoretical noise values and the circuit's transfer function. The noise should be well below the actual displacement noise, and I believe the TF looks good for actuating active damping. 

Finally, this morning, I've hooked the whole thing up, and gotten traces that show the differentiating happening.

However, when I connect the damping signal to the coil actuator, the ringdown time of the blade doesn't change. I think I don't have enough gain to drive the coils. I will insert a BUF634 and some gain to push more current this afternoon. Here's a trace of the blade ringing down, and the differentiator signal. 

 We started with the blades damped with rubber + magnets (6.2 sec 1/e time, blue trace is output of the shadow displacement sensor):

Then, I attached the shadow sensor and differentiating circuit, with a bit more gain than this morning (3.2sec):

Looks good, so I cranked the gain way up! (.54sec):

Even better! Next, I took out all of the passive damping. This was a giant pain, and it'll be awhile before the michelson is aligned again, because I really had to move things around. 

With no damping, the blade oscillations look like this:

Turning on the active damping circuit, and....

By eye, looks like about a 3 second 1/e time. So, already, the active damping is more effective than the passive damping. 

Next up: hard-wiring some permanent circuitry for all this. 

 Long radio silence due to cymac work. Came back to crackleland, and have been beating my head against alignment woes. Today I have achieved fringes on both ports! 

This was finally possible by taking the plate out of the can and putting everything together from scratch. Because I am outside of the can, the cables for shadow sensors + actuators don't reach. Thus, I couldn't damp, and so the fringes are sporadic, since the masses are free to swing. Nevertheless, in the absence of a squishy wobbly table, I was able to align both ports.

Pictures of the layout and scope showing some fringing. (Ch1 is antisymmetric port PD, Ch2 is symmetric port PD)

Procedure looked like this:

• Take everything out of the can
• Use iris, fiber output and two steering mirrors to produce a level horizontal beam. 
• Use 45 degree inclined mirror and improvised plumb bob to produce vertical beam. 
• Put end mirror (on the bottom of a mass) into vertical beam. Retroreflection consistent across different rotations == level end mirror. Repeat for other end mirror.
• Arrange first mass + inclined mirror to get acceptable wedge angle that will fit in BS cube. Use iris to ensure inclined mirror is at right pitch angle. 
• Position cube. Use iris to confirm its leveling. (Two reflected beams are coming out of the cube, two tilt degrees of freedom.)
• Position second mass + inclined mirror, adjust BS rotation as neccesary. Use iris to fix mirror pitch.
• From here on out, I touched second inclined mirror and BS only as my degrees of freedom. Pitch should be mostly aligned from iris use.
• Align by touching 2nd inclined mirror yaw and BS rotation, and looking at near field (card near AS port) and far field (letting Sym port shine across the room)
• Position PDs, tweak alignment until fringes pop up. 

Next steps:

• Recalibrate shadow sensors, since I repaired one a little while back. 
• Unplug PDs. Arrange a clever way to lift the plate into the can without destroying everything...
• Damp masses, align to good contrast. 
• Lock! (Either with slight modification of old control circuit, or bring a cymac over.)

In cymac land, I'm facing some trouble retrieving written frames, but the DAC and ADC seem fine, but this is on Jamie's prototype that will stay at the 40m. The future 050 cymac doesnt' have ADC or DAC cards in it yet, but it is all set up to receive them, but has the same frame reading issue. I will continue to work on these in parallel with crackle work, but it'll probably be more productive to modify my old circuit for the immediate future before completely jumping ship to cymac control. 

Nic and Dmass helped me lift the plate into the chamber, but I unwisely had some things hanging over the plate's edge, which meant the whole thing didn't fit...

Came in yesterday, changed the layout slightly, aligned outside, lifted in, touched up the alignment, saw a little bit of fringing. Then, I placed the coils + shadow sensors and turned on damping. Better fringes were achieved: 

Tomorrow, I will take some time to align with good contrast, and then I hope to be able to lock!

With Koji's help, I was able to lock the interferometer. Noise level is compared to the last lock in the following plot:

I'm not sure what's happening at ~100Hz, and am perplexed to see the high frequency noise be not much different. Koji made a variety of suggestions of what I should do to improve the performance of the servo, which will be my current focus. 

 I used the following line to change your PDF into PDF/A format so that the ELOG could make a thumbnail of it:

gs -dPDFA -dBATCH -dNOPAUSE -sDEVICE=pdfwrite -sOutputFile=out-a.pdf lockcomp.pdf

I injected a common mode line at 24 Hz (1000 cts.) and looked at the Michelson error signal. The line was very well visible. I changed the blade B driving gain from -1.28 to -1.37 to reduce the leakage by at least a factor 20, see plot:

blue = reference no line
green = before balancing
red = after balancing

I ran the demodulation analysis (same as the one for mich signal) on the current monitor measurements. I checked that the right channel is now the DIFF not the SUM.

Using crackle_2017_05_06 log with over 100 segments data, we can see what modulation components and amplitude we get, in the figure below on the left two columns:

As expected we have large +2FI components in almost all frequency bands. The important 1F components are finite but negative. I am presenting the MICH demodulation results on the right two columns. We don't see much demodulated components as we see in the current monitor signal demodulation. This indicates that the mich-projected level of actuation noise is small. It would be still interesting to project the modulations found in the actuation noise into MICH using the actuation model and compare them with the modulations measured in MICH. Probably, we will subtract out the actuation noise modulation confidence intervals from the MICH ones, and see what’s left.

Using the actuation model same as the one used for mich calibration, we can project the current monitor multi-band demodulation student-t test results to the mich spectrum, and compare directly with the mich demodulation student-t test results. What I am doing right now is to use the 10th order open-loop transfer function zpk fit, calculate the frequency response to the mean frequency values of each bands, and scale the current monitor demodulation in each bands correspondingly by the magnitude of the actuation transfer function (uN2mW), and multiply by the nominal optical gain (mW2m).

The figures below are projection results for [2017_05_06, f_drive = 0.0633 Hz, Amp = 16000 cts] test run. The current monitor demodulation results are shown in dashed-line confidence interval bars (green - / red +) and upper limit lines (black), superposed to the mich demodulation results solid-line confidence interval bars and upper limit line. The results match reasonably well in orders of magnitude, but it seems that a factor of two is missed -- I am showing the current monitor projection with additional scaling of 1/2 in the right figure below.

I attach the sample analysis code + the demodulation results for 2017_05_06 run to this elog. Simply run current_projection.m in the folder will generate the projection plot.

I think I know where the factor of two is coming from.

The actuator transfer function you used, which is the one implemented in the calibration, gives the MICH motion divided by the LOCK signal. But the lock signal is then sent with equal gain to two coils, so it actually is a measurement of the sum of the Z1 and Z2 actuators.

The noise measured in the difference of the two current monitors gives the incoherent sum of the contributions coming from the two actuation chains. It already includes the right way to combine the contributions of both Z1 and Z2. Threfore it must be multiplied by the response to ONE single coil actuation, which is the calibration TF divided by two.

Using the mich demodulation and the actuation noise demodulation projection method described in Elog 1722 and 1723, I analyzed two different amplitude driving crackling measurements: A_drive = [16000, 32000] counts. For each driving amplitude, we have three different frequency driving tests: F_drive = [0.0317, 0.0633, 0.095] Hz.

The odd pages (pg1, 3) show superimposed mich and currmon projection results. Using error propagation analysis I am able to subtract the current monitor demodulation projection results from the mich demodulation results in mich spectrum. The subtracted demodulation results are shown in even pages (pg2, 4).

In general, larger amplitude driving and lower driving frequency gives higher [+1FI, +1FQ, -2FI, -2FQ, +4FQ] demodulated noise in low frequency regime (10-60 Hz). I think except two sets of tests behave off-the-place:

1. 2017_05_06 [0.0317, 16] test has significant negative 1FI demod noise.

2. 2017_05_30 [0.095, 32] test seems to exhibit higher [+1FI, -2FQ] demod noise than its lower frequency driving counterparts (with the same amplitude driving).

However it's not very straightforward to compare the demodulation results between tests. I will try to integrate the demodulation results together into single plots for a quantitative comparison. I also want to try subtracting the projected currmon demod results segment by segment and do student-t test afterwards, just to confirm the rsults agree with the error propagation method.

The figure below illustrates how I plan to integrate multiple-test results into single demodulation plots. The plot only takes the measurement runs that exhibit positive demodulated (mich-currmon) in low frequency regime (10 - 60 Hz) into account, to avoid the plots go too busy. I want to make some further improvement on the plots, and will include the negative demodulated amplitude results soon. 

Another possible way of presenting the analysis results, is to integrate the demodulation noise power over low frequency regime (select 10 - 60 Hz for example) and plot versus different frequency driving, with two curves representing the two different driving amplitude sets of runs. Still working on it...

The plot below shows the Michelson locking signal and the measured actuation noise (using the coil current monitors). The traces are taken in two configurations: in both cases the Michelson interferometer is locked, but in one case the low frequency common mode excitation was off, while in the second case it's on (f = 63.3 mHz, amplitude = 16mN).

Clearly, there is more actuation noise when the low frequency drive is on. 

Using the current actuator model, I can project this noise into the Michelson sensitivity. The result is shown below, in the two same configurations discussed above

Actuation noise is normally quite close to the sensitivity, and actually limiting below 20 Hz when the common drive is on. The projection matches perfectly the measured Michelson spectrum. As pointed out in other elog entries, when the common mode excitation is on, the actuation noise is non stationary

The coil monitors filter banks include a model of the actuation, to calibrate their output in MICH meters. 

The nasty slope between 10 - 40 Hz seems to be related to a touching issue. After recovering alignment today I found the extra noise in 10 to 40 Hz noise transient.

In time series of the error signal, we can clearly see glitches of noises when this noise arises. I noted also that the fringes are fast too, and it usually takes multiple tries to lock the Mich with ~ 0 actuation. I thus suspected a touching issue, so I tried leveling the table carefully to a configuration with fringes running slow. Then the 10 - 40 Hz noise behavior is improved and stabilized. I uploaded the newly calibrated actuation function for current monitor and the measured DAC noise matches the MICH spectrum in the range ~ 15 - 35 Hz. 

Quote:

The coil monitors filter banks include a model of the actuation, to calibrate their output in MICH meters.  However, this calibration is old… We should update it with your new measurements. Using this old calibration, DAC noise is very close to the MICH spectrum, see the plot below. Maybe with the updated calibration of the actuation it will match.     We'll have to check if the current monitors are working properly: I mean, where their sensing noise is. It’s possible that the orange trace is all coil current monitor sensing noise  See also my old elog https://nodus.ligo.caltech.edu:8081/SUS_Lab/1482 for a suggestion on how to change the Z coil whitening. Also, it looks like the splitting between Z*_LF and Z*_HF could be improved. Finally, a lot of the correction comes from the ~9 Hz peaks. Maybe it would be good to try to understand where that comes from and try to damp it

As I suspected, the coil current monitors are limited by sensing noise. The plot below shows that the SUM channel is the same with the Michelson locked or unlocked. So the estimate below of actuation noise might be wrong. Still worth checking by changing the series resistor or better the whitening

Federico, Gabriele, Xiayoue

Lots of works done today!

1) We found some larger and stronger magnets (Electron Energy Corporation, dia. 1/4 inch, lenght 1/2 inch, material Sm-Co), useful for bigger force in blade actuation

2) We found and old OSEM support with a larger hole and a wounded coil, perfect match with the bigger magnet

3) We have populated one more OSEM electronic board (to be used only as displacement sensor)

4) We found a (very old) Virgo Coil-driver electronic; checked, it works perfectly

Specs are: bandwidth DC-10kHz, Gain = 10, max out voltage 40V pk-pk, max out current 4A pk-pk
(BTW: how a *20 years old* Virgo electronics has reached the west Bridge Crackling Noise laboratory? Mistery!)

We've made some actuation test with a non loaded blade (larger magnet glued on top end, smaller magnet simply attached on the bottom end and used as shadow-meter): it works perfectly.

We then have loaded the blade with 5.2kg (~25% of max load), obtaining a resonance frequency of about 1.5Hz

We have redo the actuating test: according to the OSEM calibration, 8Vpp on the 15 ohm coil (~0.5A pk-pk) gives a displacement of ~350um

All the setup is now ready for further tests (see attached photos)

I came back to add whitened signals to disk. In order to make and install the new model I broke the lock.

On the side I found the Michelson already got misaligned a bit. I realigned to the maximum visibility, and started the lock from 10 pm. The pressure went up to 518 mTorr.

DC motors - Tuning Adjustable Resistors

I first tuned the resistors of our DC motors circuits, such that the voltage meters read close to 0V when the strain gauge sensors are not stretched by the plate. A zero voltage reading would later help us know when the plate is at equilibrium. The equilibrium is unstable and so the plate moves either up or down. I tuned the resistors for the bottom sensors while the plate was stuck against the top ones and vice versa. The BE adjustable resistor (R2 in the picture) was very sensitive and acting strangely; the voltage reading would change as long as the screw driver touched the resistor, making it impossible to know whether we are close to 0V while adjusting. The resistances also drifted away from their values throughout the day. Whereas the initial offset was set within 5-10mV, at the end of the day it had grown much larger as is evident in the following image:    

Bode Plots in Matlab

I developed a code in Matlab to read .dat files and create Body plots. For future reference, Matlab files should not include a "." in their name, since Matlab recognized whatever comes after the period as the extension of the file. Following is the code for the phase plot and the results:

>> cd('C:\Users\Γιώργος\Dropbox\maglev\SURF\progress_report\Transfer Functios of Coils')
>> load 008ASC.dat
>> G = semilogx(X008ASC(:,1), X008ASC(:,2))
>> axis ([0.5 1000 -160 0])
colorss = {[0.5  0.5  0.5],
           [0.8  0.3  0.7],
           [0.0  0.0  1.0],
           [0.97  0.1  0.0],
           [0.1  0.9  1.0],
           [0.2  0.8  0.1],
           [0.4  0.4  1.0]};
for k = 1:length(G)
    set(G(k), 'Color', colorss{k});
end
grid
grid minor
%axis tight
hXLabel = xlabel('Frequency [Hz]');
hYLabel = ylabel('Phase [degrees]');
title('Phase of the Low-Pass filter', 'FontWeight', 'bold', 'FontSize',12)
set( gca                       , ...
    'FontName'   , 'Times'     , ...
    'FontSize'   , 15          );
set([hXLabel, hYLabel], ...
    'FontName'   , 'Times',...
    'FontSize'   , 15          );
set([Legend, gca]             , ...
    'FontSize'   , 15          );
%set( hTitle                    , ...
%    'FontSize'   , 12          , ...
%    'FontWeight' , 'bold'      );

set(gca, ...
  'Box'         , 'on'     , ...
  'TickDir'     , 'in'     , ...
  'TickLength'  , [.02 .0] , ...
  'XMinorTick'  , 'on'      , ...
  'YMinorTick'  , 'on'      , ...
  'YGrid'       , 'on'      , ...
  'XColor'      , .1*[.3 .3 .3], ...
  'YColor'      , .1*[.3 .3 .3], ...
  'FontSize'    , 25, ...
  'LineWidth'   , 1.5        );

 HE boards and coils test

I also measured the output of the HE sensors to see whether there were working fine:

AC1=105mv, AC2=128mV, AC3=137mV, S1=1.73V, S2=82mV, N=16mV, W=7mV. In the same way, I measured the resistances of the coils (since we have not yet created a signal with the computer) and they worked fine, too (same values as before, slightly higher resistances, possibly because of the long ribon wires attached).

Tomorrow, I will start working on Simulink and learn how to use the computer to provide the feedback filter.

Today, I adjusted the filter to control the Michelson.
After this operation, I could control the Michelson for several second. (Graph)
Following is the filter setup.
Shadow PD -> SR560(BP=10Hz, g=5) --------------------------> SR560(A-B, g=1) -> circuit(g=10) -> coil
Michelson PD ---> SR560(A-B, LP=300Hz, g=1) ------> SR560(A-B, g=2) ----^
                        |            ^----function_generator(0.07V)         ^
                        |-> SR560(BP=100Hz&300Hz, g=5, inv) --------|
I will write this setup legibly when the configration is decided.

At about 3pm I vented and substituted the pump with the smaller one (it's the one we were using for Crackle1).

I also leveled again the table (all legs at about 80 psi, which is indicated on each leg as the maximum rating). I took some time to move and re-center all OSEMs. Not clear which one was touching, but I ended up taking all out and putting them back in one by one.

At about 4:55pm I switched off all damping and left the lab, to collect some data and check that the board and blocks are free. The following spectra show that everythign is free

Chamber close at 5:30pm, pump down started. AT 6:00pm the pressure is 1.5 Tor and decreasing.

IFO locked for most of the night, although with not very good sensitivity. There is large coherence with the QPD signals, so I suspected a larger than usual coupling to beam jitter. I wanted to measure the coupling of beam jitter to MICH by injecting som beam jitter

2:10pm pumped the back left leg, plumb line is centered
2:15pm Michelson locked, but the sensitivity is horrible, even worse than during the night
2:30pm aligned the PDs by 1) misaligning the Michelson to reduce fringes 2) scanning the X and Y alignment of the input beam to find the center on the PDs 3) fine tune by centering the QPD
2:33pm Michelson re-aligned and locked, sensitivity is still very bad. Leaving it locked for a while to investigate
3:32pm all damping loops off, IFO unlocked
5:10pm checked that the OSEM signals are good, so it doesn't look like something is touching
5:50pm IFO locked

2:40pm chamber vented
2:45pm pumped all the table legs to 100 psi, the plumb line is centered
2:45pm opened the chamber
3:05pm moved all suspension OSEMs out, move the motorized counterweight to equalize X1 and X2 (both are at ~35 um)
3:10pm moved OSEM A to the right
3:15pm OSEM B was ok
3:20pm move OSEM C up and to the right
3:25pm OSEM D was ok, put back E in position (I took it off, so I don't know if it was ok before)
3:25pm put back OSEM F, recentered as above
3:25pm all suspension damping loops on, fine tuning of the OSEM values around 0
3:40pm all OSEMs are in a good position, but X1/X2 have shifted, so I moved the couterweight and recentered all OSEM again
3:45pm moved OSEM B to the right, moved D and E to the right
3:50pm moved OSEM C down and OSEM F to the left
3:55pm all OSEMs are good but X1/X2 has shifted again. I moved the counterweight to equate X1 and X2.
             Moved F to zero its value. All suspension OSEMs are very close to zero, except for D and E which are close to 200um as desired
3:57pm laser is on, recentered the QPD by moving the input mirrors
4:00pm the QPD is centered
4:03pm Michelson aligned, good fringes
4:05pm all damping loops are off, taking some data to check if the system is free
4:20pm checked all OSEM spectra, all look ok. Laser off
4:25pm checked that all coils are working
4:25pm as a test, I put the acoustic enclosure on the table. As shown in the attached plot, only OSEM A shifts by 100 um
4:40pm chamber closed, pumping down without clamps. Pressure is going down
4:50pm pressure is at 60 torr, still going down
5:00pm pressure is at 8 torr, still going down. Leaving the pump on overnight.

First attached plot shows that when the bell jar is in position, there is no large change in the OSEM values.

Second attached plot shows the change in OSEM values when the acoustic enclosure is put on the table and taken off.

The third plot shows a trend of all OSEM during pump down.

Checked all OSEMs:

• moved A to the left (it was very close to touching)
• B looked ok
• C was too high, but magnet was loose and it came off. I reglued it, and resoldered all OSEM wires (three broke). I added heat shrink on all connections. I also rotated the mount, so that now the OSEM is fully inside the restraining ring and the nylon set screw can be used to secure it
• D was close in the direction perpendicular to the borad, moved it
• E, same as D
• F was good
• X1, X2 are good
• moved Y1 to the left, it was very close to touching
• Y2 was ok
• Z1 and Z2 were good

Checked that all SUSP coils were working properly. Checked also all block OSEMs. I had to change sign to X2, why?

Then I saw glitches in Z2: I traced them down to a 2.25 MHz oscillation in the transipendance stage. Actually, it was due to a poor connection with the photodiode.

I also increased the supply voltage to 16 V, since I measured only 14 V at the opamps: clearly there is some voltage drop with the power supply cable.

Suspension OSEM spectra look good, as far as we can tell nothing is touching.

After puttind back the bell jar, we didn't see a large change in the suspension signal, so I guess yesterday something bad happened. We'll have to check the trend after the pumpdown is finished.

Pump down started at about 5:30pm LT.

Aggregating all good measurements for MG, FQ based on the criteria for "good" discussed in Elog 950, we plot to have visual comparison for dynamic modulus amongst the three samples tested so far. From a rough look, the loss modulus of amorphous materials (FQ, MG) are more "linear" than that of Cu in static stress sweep. Also notice that there seems to be a consistent increase in storage modulus in all samples. From the basic linear fitting, amplitude measurements of FQ, MG, Cu_saturate all have a slope of ~ 0.15 e3. If we understand why "good" is good, and taking enough measurements on FQ to confirm the consistency, we might be able to use it as a calibration line.

 Vacuum could get down to 65 mTorr today, roughly consistent with the last time we were sealed up. 

Spent a long while coaxing the fiber coupler into cooperation, but I now have more light going through the fiber than I ever had before (~5V pd signal, with 40db gain. More should be possible, but the coupler gets very sensitive and exhibits some hysteresis, so I've left it where it is for now).

I aligned the shadow sensors and spent ~1hr trying to align the Michelson, but wasn't able to get much contrast. However, I would see much bigger contrast when the masses were swinging around from disturbances; I think I may have to tweak the mirrors on the masses, which I wanted to do anyways to ensure a nice vertical beam. Altogether, these alignment issues we have to go through every time something moves are a huge time sink; It would be nice to have some externally controlled mirrors.

For now, I'm redoing the blade transfer function measurements, to serve as a reference for changes that will be made to the blade mechanics (adding weight, shifting suspension point, etc). Ideally, we want to do this measurement with the Michelson, but using the shadow sensors will have to do for now, since we're not aligned / able to lock.

Got sick. Got marginally better. Spent many hours trying to align. Consulted with Dmass, got some tips and got closer, but not there yet. 

The shakiness of the platform and inability to damp swinging modes or lock anything down continues to make things frustrating.