I using Levin's method (Internal thermal noise in the LIGO test masses: A direct approach) of using the suceptability of a mirror to a force resembling the laser beam profile to model the thermal noise, I modeled the GQuEST end mirrors as 2 mm thick, 1 in side length squares made of silicon with a Q of 10^6. I added a coating 20 um thick made of silicon as well but with a Q of 10^3. I constrained the 4 side of the mirror to only move in the z-axis. The spot radius is 2 mm. The face of the mirror had 52 x 52 mesh points, the substrate had 10 layers of mesh, and the coating had 3 layers of mesh. I simulated from 500 kHz to 16.5 MHz with 100 kHz frequency spacing. It look 5 hours and 17 minutes. Attached is the data converted into an ASD. The coating thermal noise, due to its low Q and that the force is applied on it, dominates the "bottom of the bucket".

[Torrey, Daniel]

We set up a Michelson Interferometer using a Flexure Mirror Mount and a standard tip-tilt mirror mount as end mirrors to test if there was a resonance that produced the noise source around 520 Hz that is observed in the output filter cavity. The beamsplitter was a 50:50 non-polarizing plate, and the readout was done by a Thorlabs PDA50B2 amplified photodetector. The alignment was such that the contrast from hitting the table was above 80%. On the frequency trace, there was no discernable increase in noise around 520 Hz (see photos). I also played sinusoids at 520 Hz and did not observe any increase in the noise. This indicates this noise is not inherent to all flexure mounts. Perhaps some flexure mounts are manufactuered or mounted sufficiently differently as to produce a 520 Hz noise source. The large peaks appear to be 60 Hz noise and its overtones.

[Jeff, Torrey]

This is a continuation from 11423. We set to improve upon our characterization of how well pushing a viton ring against the mirror (M2) of the cavity damped out the mystery 520 Hz resonance. 

We first drove sound from a phone at 490 Hz, 520 Hz, 550 Hz with and without pushing on the mirror, seen in driving_3_lines.png. The amplitude of these oscilations seem to be damped ~15 dB. There seems to be an additional peak around 450 Hz that is more broadband that should be investigated.

Jeff had an idea to play some white noise and try the same thing. Result can be found in white_noise_injection.png. Here we can see the resonance is completely damped out. It should be noted that the amplitude of the white noise the phone app I was using was quite small, so this is probably just damping out the oscillations due to the ambient noise of the room. Daniel found a louder app, will repeat this with bigger white noise. We should buy a speaker and microphone for future tests. Data for these can be found in the zip.

I find this result quite strange as we are only damping one mirror. I am going to repeat this process doing the same thing to M4. 

[Ian, Daniel]

We went to the Caltech VWR Stockroom to get solvents. We got

• 1101 Acetone 99.5% ACS 4L
• 1133 Isopropyl alcohol 99% ACS: 4L
• Water: 4L

To get them we went to the stockroom linked above, filled out the PTA number, and loaded them onto the cart.

We still need to buy funnels to fill our smaller bottles. we need to buy appropriate funnels such that they are safe for these chemicals. We have a bottle for Methanol but we decided to hold off on purchasing that until later as we don't really need it yet. 

We need to move them to a chemical cabinet. Right now they are stored in a fume hood that is not turned on I want to temporarily store them in one of the chemical cabinets (see image) in one of the Adhikari labs. I want this to happen today so I will ask for space ASAP.

Another minor step towards pynq on the RedPitaya control system.  Here is the result of uploading an arbitrary wave form, playing it, and reading it back.  This is the demo described here:  https://github.com/dspsandbox/FPGA-Notes-for-Scientists/wiki/DMA-transfer

One vhdl bug had to be fixed, putting a 0 value in quotes to help the compiler.  Vivado compilers are more careful now about types.

[Daniel, Mai ,Torrey]

The leading theory on the ~520 Hz resonance seen in the spectrum of the cavity error signal is some part of the flexture mount mirrors resonating. To test this theory we pushed on the mirror with viton to see if the resonance is damped/moved at all. Our initial test has a viton O ring pushing on M2 of the cavity only. We did this with and without pushing on the mirror, results seen here image.png. We are skeptical of these results and tomorrow will:

a) repeat this test several times. It is time consuming as the force on the cavity mirrors misaligns the cavity.

b) instead of relying on the ambient noise of the room to ring up, use phones to drive at say 520 +/- 20 Hz at constant locations between tests.

[Torrey, Lee, Daniel]

After some investigation we have found a mystery resonance around ~520 Hz. IMG_0015.png is the set up used to take a spectrum of the error signal and the control signal. As seen from mysteryresonance.png (blue is error, orange control) you can see a resonance peak at ~520. This exibits resonance type behavior, i.e. drive a sound at 520 Hz and measure the level of excitation and compare it to frequencies around this, say +/- 20 hz it is much reduced. 

Additionally we found some source of phase delay from our last post by taking a transfer function accros the piezo controller, as seen in IMG_0020.png. Lee mentioned it would be useful to create a filter that inverts this amplitude loss.

The same sd card in a different RedPitaya and power supply works.  

Yellow is the input signal to the RedPitays.  Blue is the output, with offset register set to 100 mV.  The delay seems to be 100 ns.  The gain seems to be a bit more than 1.

So, we can read and write analog signals and control registers with pynq.

[Torrey, Daniel]

We aligned light from the misc category into a fiber on the power distribution sled. Instead of the 5 axis mount, we used a simple 2 axis mirror mount. We also tuned up the alignment into the other 3 fibers to maximize the power. We measured the power going into the fiber and out of the fiber to see if and how it drifts over time. Potentially, the 5 axis mounts will drift more than the 2 axis mount.

I built the example analog feedback from https://github.com/dspsandbox/Pynq-Redpitaya-125

On one RedPitaya board the firmare loads, but setting the offset value via pynq causes the board to freeze up.  Identical firmaware and test on a second board succeeds.  Now will diagnose the "bad" board.  First suspect is the power supply.

[Lee, Torrey, Daniel]

On Friday, we adjusted the loop shaping on the filter cavity to try and make the lock better. With a 1/f loop, it will lock, but goes between flickering out of lock, to oscillating as the gain is increased.

The new shaping uses the "slow" output of the "laser lock box." It shapes the fast output to have a 10Hz - 200Hz boost (pole - zero), giving 20x gain at low frequencies. The overall 1/f is then put at 1Hz in the slow path. We are able to get a UGF of 1kHz, with 30-40deg remaining. We lose 90deg by 2kHz from either the actuator or the Moku, so 1kHz is around the max UGF we can currently get.

Even with that loop, we get dropouts while talking loudly or while the fans are on. Problematic. That mean the loop can supress the peak length noise below 1.5um / 2  / (2 * finesse). Our current input coupler is 1%, so our finesse is around 2pi / 0.01 = 600 (it might be as low as 300, depending on the losses in the cavity). So our peaks/RMS are around 0.5nm - 1nm of noise.

+1 to the piezo transfer teams. both 11380 and 11373 have the transfer function data along with the plots. It looks like all of the phase we are losing in the cavity lock is from the piezo (though could still be from Moku). We'll need to fit and at least slightly invert this actuator transfer function to make a loop above 1kHz UGF. The fit can help tell us how much of the phase is from the piezo roll-off.

We are going to need it - the current loop & cavity  has marginally too much noise, with an RMS around the linewidth. That is going to be a problem once we increase the finesse of the cavity. So we'll need to reduce the noise and/or improve the loop a respectable amount.

our voices are driving 100-300Hz, so we need gain there. If we can get the UGF up as high as 3kHz, then we should also be able to move the boosts up to 600Hz or so. That should give 9x more loop gain and cover much of the need.

options:

frequency lock - good for acquiring and transitioning to aggressive lock, but not an option for operating the experiment

optimal control - we can determine how good we can do with what we have

fitting/inverting actuator - probably needed to get UGF higher. May vary with time in bad ways

sound attenuation or active noise cancellation - may be possible and should look into engineering this.

We should determine if the sound driving is from pushing the mirrors or from the index fluctuation as the pressure modulates. 11290 has calculations about the pressure pushing on a 1kg mass. My worry is that the small alignment adjust set-screws in the MKS flexture mounts are fairly compliant. Could sound drive them 1nm?
My guess is that the light going through 2m of air is the culprit.

The sound pressure of 40dB should be around 2000uP. Air pressure is 101kPa. The index of air is 1.000293

2m / (1.000293**2) * 2000e-6Pa / 101kPa * (1.000293 - 1) = 1.2e-11m

So that is a little small, but if our sound is at 80db, which would be quite loud, then it would be enough to cause this.

For the mirror motion, the mirror mounts are about 1.5in square. On the pressure level that is

(.0254m * 1.5)**2 * 2000e-6Pa = 3e-6N

of the mirror+panel is about 35g (very rough estimate). For a free mass motion that is

1 / ((2 * np.pi * 100Hz)**2 * 0.03534291735288516kg) * 3e-6N = 2e-10m which is getting in the ballpark with 60db of noise rather than 40db. The mirrors aren't free though.

and now I'm curious how that is against the spring constant of the tiny screw.

[Torrey, Daniel]

We measured the reflectivity of 775 nm light on the Newport 10D20DM.8 (1550 nm HR mirror). At a 45° AOI, there is roughly a 40-50% reflection. +/- around 5° away from 45°, the reflection drops off sharply. Thus, these mirrors cannot be used to make a bowtie cavity for 775 nm light.

Low AOI: 3% reflection, 70% "transmission" but definitely a lot of this is scattered light

45°: 40-50% reflection depending on polarization, ~40% "transmission" but definitely a lot of this is scattered light

The frame is assembled.  Custom-machined parts for the walls and inside shields are on order.