Installed the etched disk: using manually the centering ring allowed me to get the beam on the QPD. A couple of taps to the disk were enough to get the beam centered.
Pump down started at 8:52am
I opened the chamber and took the etched disk out. Inspection of the electrostatic drive does not show any sign of burn or damage.
So it seems that the problem we had previously was due to contamination of the chamber (in the first case) or of the ESD (in the second case)
I moved the roughing pump out of the clean room, adding an extension hose.
This reduced a lot the vibrations induced by the pump. In the past when the pump was running we often saw very large noise, see the red trace in the figure below. Now, in the same conditions, we get the blue trace, which is much better.
The plot below shows a comparison of different configurations:
We are quite close to the pump off condition.
I looked into a couple of turbo pump switch on periods, and in both cases when the pump speed hits 58 Hz, a resonance is excited. I'm not sure what's resonating.
Today I let the roughing pump reduce the pressure to a level lower than usual, and this seems to mitigate the effect. The disk wasn't shaking as much as it did in previous pump-downs.
I added three 1"x1" viton pads below the base plate, and realigned the entire optical system to the horizontal reference
The pads seem to reduce a bit the vibrations in the QPD_X direction, but significantly improve the situation in the QPD_Y direction, see below:
Here's a trend of the QPD signals when the IGM was turned on:
Turning it off does not bring the disk back.
The new chamber was build. The first attempt to pump down was unsuccessful because of dislocated lid. After the lit was placed properly, the chamber was pumped down to 0.1 torr within 8 min. In future a sky hook will be used to help placing the lid properly. The clamps should not be placed on the lid before the roughing pump turned on.
The turbo pump controller failed. Error code 698 – call the vendor.
The SkyHook has been put in place and bolted down to the floor.
Turbo pump controller (new chamber) was configured. Need to reduce the frequency or setup a standby mode. First pump down: E-7 range reached within about an hour. See plot: blue - old crime chamber, pink - new crime chamber.
This afternoon we started the pump-down with all the system installed into the chamber. Unfortunately the IGM vacuum gauge isn't working, so we can't be sure what the pressure is. To be fixed
I changed the set point of the test chamber turbo pump to 666 Hz. This was done by setting the "standby rotation speed" to 80% and enabling the standby condition.
I failed to interface in a reliable way the datalogger with Linux. So i hooked up the four analog outputs of the vacuum gauge controller to the fast ADC. The voltages are read and saved to disk at a reduced rate (16 Hz, throught the epics channels):
A python script running on the workstation (~/pycrime/automation_scripts/vacuum.py) reads the voltage every second and update four epics channels:
X3:CR0-VACUUM_CG0 = convection gauge of test chamber (high pressure gauge)
X3:CR0-VACUUM_IGM0 = ionization gauge of test chamber (low pressure gauge)
X3:CR0-VACUUM_CG1 = convection gauge of main chamber (high pressure gauge)
X3:CR0-VACUUM_IGM1 = ionization gauge of main chamber (low pressure gauge)
See below for an example during pump down (the test chamber is at a too high pressure for the IGM to work, so it returns a bogus number).
Here's an example of venting and pumping down
After venting the vacum chamber (CR14) a few times, checking for leaks and trying to tune settings to the gauges controller, I gave up. I removed the low pressure gauge from the newer vaccum system (CR14). Inspection did now show any obviouse depositions around the electrode (due to some burns). I will pack the gauge ans send it to the manufacture for an RMA. Took the same gause from the older vaccum system (CR0) and installed it on CR14. Started pumping down. The low pressure gause turned on just fine. Will check the preassure in an our before starting a measurement.
The two spectra below show basically no difference (blue roughing pump on, red, roughing pump off)
Instead, below is another comparison: blue same as before, standard condition, red with one of the two clean air filters momentarily off. There is some clear improvement. The second filter is too hard to switch off!
This morning I installed temporarily a second QPD to monitor the input beam. The goal was to understand where the vibrations at frequencies below 2kHz couple from. As shown in the photo, the second QPD was close to the first one.
The signals in the two QPDs were quite different, and the coherence between them wasn't great. So I concluded that the main coupling path is not through input beam of QPD vibration, but more likely real motion of the disk.
I removed the additional QPD and restored the setup to its nominal configuration. The readout infrastructure is still in the model.
The plot below compares the QPD signal spectra in different configurations (roughing pump on/off, air on/off).
The noise below ~<2kHz is making very hard to measure the Q of the first mode at 1100 Hz
The main source of noise are the clean air filters. I switched them to minimum for the moment being.
In normal conditions the RMS of the QPD signals is dominated by the 58 Hz line generated by the roughing pump. Also, when the modes are excited, they exhibit large sidebands at +- 58 Hz that are an annoyance for the analysis.
I improved a bit the level of the 58 Hz in the QPD signals by putting the roughing pump on top of a "Very Useful Box":
Despite the fact that this advanced vibration isolation is already a little bit effective, it might be good to try to build some better suspension and maybe add an acoustic isolation around the pump.
I suspended the roughing pump with four springs. The reduction of the 58 Hz peak is similar to what I got when the pump was sitting on a box. So most of the coupling is due to acousting noise.
The wandering line I mentioned in my previous elog, which is spoiling most of the sensitivity, turns out to be power noise of the laser.
I used a Thorlabs PDA100 and a SR785 to measure the power noise out of the laser directly, and saw a huge forest of peaks above 20kHz. Among them, a couple of peaks are moving up and down in frequency very fast. The plot below compares two different times of the Thorlabs HNL210L laser (the new one, 21 mW) with the old JTSU laser we are using for the test setup:
The noise of the new laser is cleary much larger (even after the laser has been on for some time) and non stationary. This is a big issue for us. I will contact Thorlabs to inquire if this behavior is normal.
The attached video file shows the peaks dancing around on the SR785 screen.
The turbo pump of the CR0 chamber runs at 833 Hz. It causes vibrations that pollute the measurements in the CR1-4 chamber. In particular CR1 is extremely sensitive and the line is highly up-converted. It's not clean why CR1 is more sensitive than CR2-3-4.
I resolved the factor of two from Griffiths' discussion of dipoles in non-uniform electric fields. The force on a dipole in a non-uniform field is where is the difference in the field between the plus end and the minus end. Component wise, where d is a unit vector. This holds for y and z, the whole thing can also be written as . Since p=qd, we can write .
Jackson derives it differently by deriving the electrostatic energy of a dielectric from the energy of a collection of charges in free space. He then derives the change in energy of a dielectric placed in a fixed source electric field to derive that the energy density w is given by . This explicity explains the factor of two and is an interesting alternative explanation.
par.a = 75e-3/2; % radius [m]
par.h = 1.004e-3; % thickness [m]
par.E = 73.2e9; % Young's modulus [Pa]
par.nu = 0.155; % Poisson's ratio
par.rho = 2202; % density [kg/m^3]
%Calculate fundamental modes of the disk
[freqs, modes, shapes, x, y] = disk_frequencies(par, 10000, 1, 'shapes', 0.5e-3);
%Now we extract the force profile from the COMSOL model
function out = model
% Model exported on Jul 14 2017, 14:47 by COMSOL 184.108.40.2062.
model = ModelUtil.create('Model');
fpro = zeros(6, 27);
no = 1;
for count = 1:.2:2
gap = strcat(num2str(count), ' [mm]')
model = fst2(gap);
fpro(no, :)= product(:);
no = no + 1;
I am posting a diagram of the geometric parameters that I swept. The only one not included is the vertical space between the ESD and sample that sweeps perpendicularly out of the image
From the data I have gathered from a variety of MATLAB sweeps, I think that the optimal geometry I can produce has the parameters in the attached image. Neither the original or optimized drawing is to scale. The gap between the arms of the electrodes should be 1.25 mm, the arm width 0.55 mm, the arm length 16 mm, and the offset of the arms 3.5 mm.
It is also optimal to place the ESD as close to the sample disk as can reasonably be achieved, at around 0.5 mm away. Since the force on the disk scales exponentially with the distance from the ESD, decreasing that gap is the most substantial way to impact the excitation. Decreasing the gap from 1 mm to .5 mm increases the excitation of the modes by approximately a factor of 8.
From my simulations, the shift in geometry alone still has a useful impact on the excitation. Modes 1 and 3 are the only two modes that are less excited by the new geometry, mode 1 is 10% weaker and mode 5 is 5% weaker. Modes 5 and 6 are nearly unaffected by the shift, mode 5 is 2% stronger and mode 6 is 5% stronger. Modes 7, 18 and 19 are outliers, 7 is excited by a factor of 7, 18 by a factor of 4 and 19 by a factor of 17. The rest of the modes are improved by between a factor of 1.5 and 3. For mode numbers, shapes, and frequencies a plot is included.