Attachment 1: The Circuit Diagram,

>>  the TIA with a gain of 5.1k

>>  Differential Circuit with a gain of 100.

Attachment 2: Noise across TIA

The input-referred current noise across the trans-impedance amplifier was measured using SR785 and was compared against the incoherent sum of input-referred current noise graph obtained from ZERO simulation.

Attachment 3: Noise across Differential circuit

The input-referred current noise across the Differential circuit was measured using SR785 and was compared against the graph obtained from ZERO simulation.

During measurement of transfer function using SR785, a source of 1V was given to a 10k resistor which was connected in series with the circuit taken into consideration. The channels of SR785 were set to AC coupling and input was set to Ground. Apart from that Hanning window function was used for measurements from SR785.

Input Referred noise to be calculated for trans-impedance.

Attachment 1: The Setup,

The setup was made on a breadboard, as per the circuit diagram.

Attachment 2: The Circuit Diagram on paper,

>>  the TIA with a gain of 5.1k

Attachment 3: Noise across TIA on SR785 screen

This is the noise as seen on the screen of the SR785

Attachment 4: Noise across TIA captured using the SR785 template file

The input-referred current noise across the TIA was measured using SR785 and was compared against the graph obtained from ZERO simulation.

** There is a difficulty in capturing the exact form of data as displayed on the SR785. Previously, I had captured the full span by keeping the start frequency as 1Hz. This gave only two points between 1Hz and 100Hz, all the rest of 798 points out of 800 were plotted at a higher frequency. This gave a straight line of higher magnitudes at low frequency(1Hz to 100Hz) since there were only 2 points. But the plot on the SR785 screen looks different.

** This time, I divided the measurements into 3 parts, 1Hz to 100 Hz, 100Hz to 10kHz and 10kHz to 100kHz. This left me with a graph which looks clumsier than the previous ones. I guess if there is a way that the template files of SR785 can be modified then it can give a graph which will align properly to the simulated results. For the time being, I can try again to take measurements in 3 sections but this time with fewer(300 instead of 800) points.

** It's difficult to avoid 60Hz harmonics with a circuit kept in open as this one. Lots of its effects are visible in the plot.

Input Referred noise to be calculated for trans-impedance.

Attachment 1: The Circuit Diagram on paper,

>>  the TIA with a gain of 5.1k

Attachment 2: Noise across TIA 

The input-referred current noise across the TIA was measured using SR785 and was compared against the graph obtained from ZERO simulation.

** This time, I divided the measurements into 7 parts, 0-800, 800-2.4k, 2.4k-5.6k, 5.6k-12k, 12k-24.8k, 24.8k-50.4k, 50.4k-101.6k. The number of points for each was 800. Hanning Window function was used in the template file and the Input channels were grounded. 

Attachment 3: Noise across TIA 

The input-referred current noise across the TIA was measured using SR785 and was compared against the graph obtained from ZERO simulation.

** This time, I divided the measurements into 1 part, 10-6.4k. The number of points for each was 800. Hanning Window function was used in the template file and the Input channels were grounded. 

** To measure the noise the output was measured at the pin6 of the OpAmp.

** A source of 1V was applied to the circuit by keeping a 10k in series with the input of the circuit when Transfer Function was being measured.

** It's difficult to avoid 60Hz harmonics with a circuit kept in open as this one. Lots of its effects are visible in the plot.

Attachment 4: Noise of the power supply used

The power supply was observed to be used to be noisy. 

Input Referred noise is calculated for the circuit that is to be used for characterization of photodiodes. For the biasing 12V was used from SR560 as it provides cleaner voltage as compared to other voltage supplies. 

Attachment 1: The Circuit Diagram TIA

>>  the TIA with a gain of 5.1k

Attachment 2: Input Referred noise of TIA 

The input-referred current noise across the TIA was measured using SR785 and was compared against the graph obtained from ZERO simulation.

Attachment 3: Differential Circuit

>> gain of 100

Attachment 4: Input Referred noise of Differential Circuit

The input-referred voltage noise measured using SR785 and was compared against the graph obtained from ZERO simulation. 

Attachment 5: Whitening Filter Circuit

>> gain of 10

Attachment 6: Input Referred noise of Whitening Filter Circuit

The input-referred voltage noise was measured using SR785 and was compared against the graph obtained from ZERO simulation.

Some points that were observed: 

*** I am observing deviation from simulated results at higher frequencies. Presently, I am unable to understand the cause of this deviation. 

*** At low frequencies deviation from simulated results is perhaps caused due to 60Hz harmonics and 1/f noise.

A comparison between the types of voltage regulators was done in order to know which should be preferred to provide a clean bias supply.

Attachment 1: The Circuit Diagram of the voltage regulator 

Tere are 3 types of voltage regulators that were tested. 

a. LM317

b. LM7915

c. LM7815

Attachment 2: Output Voltage noise of all 3 voltage regulators

A voltage of 12V(+/-) was provided using SR560 to the respective input of the regulator IC and the output noise across each were measured using the SR785.

LM317 will be a better choice to make a voltage regulator for the circuit mentioned in elog entry 2381.

Answers:

1. For LM317 I received the output of 11.2V and for LM7915 -10V and for LM7815 at 10.4V

2. I did a mistake with the supply. Next time I won't use SR560 and will use a voltage supply instead. 

3. The diode is for protection purpose. How should I use the diode for 7915, should I put it in forward bias or not use it at all?

4. I did check the voltage supply provided by SR560 using a multimeter, they were 12V.

Referring to elog entry 2385. I did the measurements again because mistakingly I had been using the SR560 with it's AC supply on. This time I used the 12V supply with no connection to the AC supply.

Attachment 1: The Circuit Diagram TIA

>>  the TIA with a gain of 5.1k

Attachment 2: Input Referred noise of TIA 

The input-referred current noise across the TIA was measured using SR785 and was compared against the graph obtained from ZERO simulation.

Attachment 3: Differential Circuit

>> gain of 100

Attachment 4: Input Referred noise of Differential Circuit

The input-referred voltage noise measured using SR785 and was compared against the graph obtained from ZERO simulation. 

Attachment 5: Whitening Filter Circuit

>> gain of 10

Attachment 6: Input Referred noise of Whitening Filter Circuit

The input-referred voltage noise was measured using SR785 and was compared against the graph obtained from ZERO simulation.

Attachment 7: The Scripts

All the scripts and data used in the measurement.

** I did notice a reduction in 60Hz harmonics but I still see a deviation from simulated results at higher frequencies and at frequencies below 10Hz.

Referring to elog entry QIL:2387. I did the correction with the voltage supply and now provided a supply of 18V(+/-) to LM7815 and LM7915. The position of diodes was also corrected for LM7915. The Electrolytic Capacitors(100uf) I am using are getting heated when using with LM7915 only. I didn't find any tantalum capacitors of 100uf in EE shop. Should they be replaced with some other capacitors?

Attachment 1: The Circuit Diagram Voltage regulator

The component used                   Input Voltage                              Output Voltage

a. LM7915                                    -18 V                                  -15.1 V

b. LM7815                                    18 V                                    14.86 V

c. LM317                                      18 V                                    17 V      

Attachment 2: Output Voltage noise of regulator circuit

The noise observed using SR785 at the output of each regulator.

As it was observed that normal voltage supply is noisy and not suitable for our circuit, we plan to use a voltage regulator that will help us provide a clean supply. Referring to previous elog entries the corresponding corrections were made( polarity of electrolytic capacitors, ceramic cap in parallel to electrolytic, 3V difference between input and output of respective regulators).

Attachment 1: The Circuit Diagram of Voltage regulator

The component used                   Input Voltage                              Output Voltage

a. LM7915                                    -18 V                                  -15.1 V

b. LM7815                                    18 V                                    14.86 V

c. LM317                                      18 V                                   14.96 V      

Attachment 2: Output Voltage noise of regulator circuit

The noise observed using SR785 at the output of each regulator is shown. It clearly shows that LM317 manifests less noise in comparison to LM7915 and LM7815. It will be therefore a good idea to use this to provide 15V bias in our circuit.

Attachment 3: The Scripts

Find all the scripts and data used in this measurement.

The photodiode needs a 1V bias so for a clean bias we have decided to use a sallen key low pass filter with a cut off frequency at 1Hz. The quality factor of the designed sallen key filter is 0.707.

Attachment 1: The Circuit Diagram of Sallen Key filter

The gain of the circuit is 1. 

Attachment 2: The transfer function of the filter

We can see the cut off frequency at 1Hz

Attachment 3: The Input Referred Noise of filter

The input-referred voltage noise was obtained using SR785 and compared with zero simulation. It deviates a lot from the simulated results by a factor of 100.

Attachment 4: Scripts

Find all the data and scripts used for the measurements.

NOTE: Noise plot below had inconsistent units; please ignore.

The circuit has been soldered(refer entries 2399) and the noise for Sallen Key was analyzed

Attachment 1: Circuit Diagram of Sallen Key low pass filter( cut-off= 1Hz)

Attachment 2: Transfer Function of sallen key.  The Frequency response Measurement was done using the Swept Sine group. The input range was -50dBVpk.

Attachment 3: Noise comparison between zero and SR785 measurements. The noise matches the simulated results to a great extent and also it's less noisy so can successfully be used to bias the photodiodes(1V).

Attachment 4: Zip File

The 1/f noise and dark current density were analysed for Sb3513_A2 photodiode. 

Attachment 1: Dark current density plot

It was observed that the dark current density has a very less difference for measurements taken across 500um, 750um and 1000um. It means that the leakage current is of low magnitude.

Attachment 2: 1/f noise at 10Hz

The 1/f noise for 500um, 750um and 1000um was plotted and 1/f noise is high for 1000um as the bias is increased. and 1/f noise is high for 500um at low bias voltages.

Attachment 3: Zip file

The noise analysis for TIA was done. The circuit was in open but kept away from SR785 (to avoid any noise effect)

Attachment 1 and 2 show how the setup was placed. The wires were kept in a way that there is no tension. The wires that were used for connection from the voltage supply were twisted in order to avoid any inductance issue. The input range was kept at -44BVpk (this was maintained at all points when taking measurements with SR560) while using the SR785. SR560 was used with a flat gain of 100 in order to get above the noise of SR785 and also the AC coupling was used. LM317 and LM337 were used to provide a 15V(+/-) supply to OpAmp. The OpAmp used here is Op27. 

Attachment 3 shows the noise analysis across TIA(using Op27). It was observed that the voltage regulators help in noise reduction to a great extent at low frequencies but somehow at around 8kHz, a huge noise bump is being observed. I also checked the noise by using directly the voltage supply at the lab. It does impart high noise at low frequencies but it's clearly visible that noise bump at 8kHz isn't there. The noise bump exists only when the voltage regulators are being used with the OpAmp. I did check if the output of voltage regulators were oscillating due to some reason but they provided a constant output of 15.04V(+/-).  I did check if the OpAmp was broken but it isn't the case because the difference between the voltage at pin 2 and 3 is zero, I have two TIA on my board so I checked the noise for both of them and I observed the same results.

Attachment 4 shows the noise of TIA using LT1792. It was seen that the 8kHz noise bump is evident on even changing the OpAmp. 

I am unable to understand how is this issue coming up. I did the measurement quite a few times just to be sure It's not a one-time thing but the noise bump is dominant. 

Attachment 5: Zip

Noise analysis was done using SR785. SR560 was used with a flat gain of 100 to get above the noise floor of SR785. The input range was constantly maintained at -44dBVpk for all measurements. Voltage regulators LM317 and LM337 were used to power the circuit. 200 averages were taken for all the measurements. The TIA was configured with a 5.1k feedback resistor and 100pf feedback capacitor. Please refer elog:2390 for better understanding of the circuit diagram. 

** Referring to elog:2411 the 8kHz noise bump went away on its own without changing anything in the circuit. I have no clue how it happened and why it's not happening again. 

Attachment 1: Noise analysis using OP27 in transimpedance amplifier. At Frequencies below 100Hz, data was taken in 4 parts, starting from 0Hz with a span of 25Hz but with 10 number of averages(fewer averages were taken only in this case).  At high frequencies(above 100Hz) data was taken with 200 averages. A noise was observed to be 10pA/rtHz was observed at 10Hz and 3pA/rtHz above 300Hz.

Attachment 2: LT1792 was used in this case. It was seen that it is less noisy as compared to OP27. The noise was observed to be 2pA/rtHz above 20Hz.

Attachment 3: LT1012 was used for this measurement. The noise was observed to be 3pA/rtHz above 20Hz.

Attachment 4: AD820 was used for this case. The noise was observed to be 3pA/rtHz above 500Hz.

Attachment 5: OPA140 was used for the TIA during this measurement. The noise was observed to be 2pA/rtHz above 2Hz.

Attachment 6: Noise comparison between all the OpAmps used. It was seen that OP27 isn't able to deliver performance as expected because it is getting affected a lot by the noise(1/f noise). OPA140 performs better than all the others. 

Attachment 7: Zip file to re-create all data

**edited as per suggestions in elog:2420

The dark noise of IG22X2000T9(serial: X8906 and X8905), Extended InGaAs photodiodes was measured. A low pass sallen key filter(using OP27) with a gain of +1 and cut off frequency of 1Hz was used to provide the bias voltage to the photodiode. A transimpedance amplifier(using OPA140, refer elog:2416 for noise spectrum of TIA) with a gain of 5.1k was used to convert the output current of the photodiode to voltage. The input range was maintained at  -50 dBVpk during the measurement.

A bias voltage of 1.017 V was provided and the output voltage across the transimpedance amplifier was observed to be as follows:

X8906:    -0.030V, which implies that the dark current was -5.887uA.

X8905:   -0.097V, which implies that the dark current was -19.01uA.

Attachment 1: Setup representation

Attachment 2: Experimental Setup. It was made sure that the cables are free from any tension. Connections were made using BNC connectors. The transimpedance amplifier and sallen key filter were placed in a box and were not in direct contact with the optical bench. During measurement data was taken with a linewidth of 125mHz(was increased logarithmically for subsequent measurements, since measurement was taken in parts) with 200 averages for each set.

Attachment 3: Dark noise plot. The data was taken for X8906 for 5 different bias voltages. The input range was maintained at  -50 dBVpk during the measurement. It was observed that dark noise decreases with decrease in bias voltage.

Attachment 4: Dark noise plot. The data was taken for X8905 for 4 different bias voltages. The input range was maintained at  -46 dBVpk during the measurement. It was again observed that the dark noise decreases with a decrease in bias voltage.

***** The noise is observed very low for a 0V bias for both the photodiodes below 10kHz. It was observed that noise is high above 10kHz at all the bias voltages for both the series.

Attachment 5: Dark Current plot for both X8905 and X8906 series of photodiodes.

Attachment 6: Dark Current Density for both X8905 and X8906 series.

Although being made of the same material both the photodiodes have some difference in their dark current. It was observed that the photodiodes are very noisy at room temperature. I think they will deliver better performance at low temperatures. 

Attachment 7:  The 1/f noise was observed at 10Hz for both the series of photodiodes. 

Attachment 8:  Zip file to re-create the data.

**[Internal Quantum Efficiency added]

[Koji, Shalika]

Further measurements were done after elog:2419 for Quantum Efficiency of Extended InGaAs Photodiodes(X8906). A Laser of wavelength 2um was used with an incident power of 0.80+0.02mW.  The Ophir RM9 power meter was used to check the incident power and also measure the reflectivity.

Attachment 1: The Setup. A Fibre launcher was used to project the laser along with a converging lens of the focal length of 40.0 mm which was further arranged with a subsequent converging lens of  150mm focal length. A mirror was used to reflect the laser light on the photodiode at an angle of 45^o. The bias voltage was provided to pin 4 of photodiode using a Sallen Key low pass filter and the output at pin 3 of the photodiode was fed to a transimpedance amplifier (with a gain of 5.1k) which converted the photocurrent to voltage.

Attachment 2: The Quantum Efficiency is plotted with respect to different bias voltages, It was observed that the quantum efficiency increases with an increase in bias voltage. An External Quantum Efficiency of 77.4% was observed at 1V(maximum bias voltage for the photodiode). The Internal Q.E was observed to be 83.8% taking into account Reflectivity of (60.0+1) uW at an angle of 17deg. 

Attachment 3: To recreate all data

1. Start by assembling the plane mirror to the clamp. First, put an O ring inside the clamp envelope(attachment 2) and then gently place the flat mirror on that(attachment 3). Rotate it 45 degrees and bolt this setup with the metal spacer using screws and Allen Key, Follow the same procedure for another plane mirror as well. The plane mirror is 99% reflecting on one side and transmitting on the other. The reflecting surface should be placed facing inside the clamp. An easy method to find the coating of the mirror is to hold it from the sides (never touch the middle part) and then checking if the bottom surface of the mirror is visible. If the bottom part is visible, then the side facing you transmits light and hence should be towards the outside. After this stage of assembly, it will look similar to attachment 4. Note: 3/8'' screw was used for this step.

2. Next, proceed to assemble the endcap unit. The PZT should be glued centered on the endcap and the curved mirror should be glued centered on the PZT. As is it very difficult to align them properly, a jig can be used for gluing purpose. The external space has the same diameter of the PZT, the internal one has the same diameter of the curved mirror. The slots on the edges are used for the wires of the PZT. Epoxy 30 CL can be used for this purpose. A necessary support system can be assembled as per need.

In October, I drew a crude prototype of the torsional pendulum seismometer, after discussing estimated dimensions and parts with Kate (Prototype 1 Drawing.PDF) Recently, I made frame models with accurate dimensions and parts on SolidWorks. I ran through some (very slow) simulations on ANSYS for modal frequencies of the aluminum alloy frames. Our goal is to look out for any modes with a resonant frequency <100 Hz and see if there's some modification we could make to the frame to remove them.

Results: Frame with middle struts lower to the ground had highest first modal frequency of 64.318Hz. Second highest was frame with middle struts right in the middle of total height of frame with f = 64.288Hz. Lowest was middle struts higher up the frame, with f = 59.338Hz (see attached ppt)

The objective of this work is to find the frame configuration with the highest modal frequency (>100Hz) in order to provide most stable supporting structure for the seismometer.

I made multiple frame designs on SolidWorks and ran modal simulations on these drawings via ANSYS. The varied parameters of different designs include height of the horizontal structs, addition of cross bracings, presence of brackets, and struct thickness.The frames have dimensions of 3' height x 2' width x2' depth. The first six modal frequencies were considered (Fig 1). Eight different designs were evaluated (Fig 2). Their frequencies are summarized in a graph (Fig 3).

Summary of observations made:

• Having crossbeams yield 50-100Hz higher frequencies for modes 1,2,3,4.
• K crossbeams yield 10-15Hz higher frequencies than parallel crossbeams
• Thicker (60mm vs 45mm) beams yield ~10Hz higher in modes 1,2,3,4; and ~40Hz higher in modes 5,6
• Having brackets yield 13-15Hz higher in modes 1,2,3,4; and significantly more in modes 5,6

Figure 1. Six modal frequencies were taken into consideration for comparison.

Figure 2. Eight different designs were considered for comparison.

Figure 3. Different designs yielded a differing set of modal frequencies.

See attached ppt for more details.

Since modal frequencies have been calculated for different frame designs through FEA analysis, we decided to physically test the modal frequencies of the current frame in the lab.

We clamped an accelerometer onto a middle, horizontal strut of the frame. The accelerometer was connected to an amplifer, which was then connected to an oscilloscope. I banged the middle frame with my hand to give it an impulse. Without the ampifier, the amplitude of the waves was about 30mV. With the amplifier, the amplitude was on the magnitue of several volts. The oscilloscope displayed sinusoidal waves with frequencies ranging around 190-250Hz. There was also an enveloping wave with a frequency ranging around 30-50Hz. Our next plan is to transform these time domain waves into frequency domin spectra to identify modal frequencies of the frame.

Fig 1. When the frame was struct, there was a wave of 192Hz and enveloping wave of 27.8Hz (measurement made through the amplifier).

We used the signal processor to try to plot a frequency domain output of the accelerometer. The rough plot did show a noticeable peak at 73Hz and 164Hz (Fig 2). However, the signal processor is not very user friendly, and we are still trying to figure out all the settings. Hopefully we can save the file as well so that we can perform further analysis.

Fig 2. Frequency spectrum of the frame modal frequencies. The faint (lower) line is the ambient noise. The bold line is the data collected after the frame was struct with the rubber handle of a screwdriver.

Setup of the experiment are shown in Figure 3, 4.

Figure 3. The accelerometer is attached to the frame by a clamp secured by two screws. Because there is a gap in the frame, a piece of tape was used to allow more contact between the accelerometer and the frame.

Figure 4. Setup of the experiment. The frame is on the laser table, with the accelerometer mounted on the top, horizontal struct (location can vary for each test). The accelerometer is connected to an amplifier which is then connected to the oscilloscope and signal processor.

To reduce additional mass on the frame, a zip tie was used instead of the mount (Figure 5). We gathered some data, using the plastic handle of a ball driver as a source of impulse. The raw data files are attached (Vpk (log) vs Hz (linear)). The locations of testing are in Figure 6.

Figure 5. Accelerometer was attached on the frame using tape and zip tie.

Figure 6. The numbers indicate where I lightly hit the frame. The numbers correspond to raw data file name (#7,11 are ambient).

I plotted the raw data collected with the accelerometer attached to the high bar (Fig 7) and to the middle bar (Fig 8). Then I identified noticeable peaks.

Figure 7. Raw data of modal frequencies detected by an accelerometer attached to the high strut of frame.

Figure 8. Raw data of modal frequencies detected by an accelerometer attached to the middle strut of frame.

Next, Kate and I wrote a Matlab code that, given a left bound and right bound of a peak (chosen by observation), approximates a Lorentzian probability distribution in the range and fits the approximation to the raw data (using lsqfit). The function also outputs the peak frequency and the Q factor. Figure 9 displays the raw Data 8 (blue dots) and the fitted curves of identified peaks (red lines). Figure 10 displays the Data 14 and its peaks.

Figure 9. Data 8 (blue) with its fitted peaks (red).

Figure 9. Data 14 (blue) with its fitted peaks (red).

The motivation for this analysis is to observe whether the collected data are comparable to the ANSYS simulation that I previously did in order to identify the different modes present in the frame. Data 8 and Data 14 share many of the identified peak frequencies. By comparing these values with that of the simulation, many of the frequencies align with those calculated by the simulation (see attached "Peak Freq and Q factor.xlsx"). However, the collected data has so many peaks that it is difficult to discern whether the frequencies actually match the simulation frequencies or whether it's coincidence due to the large number of peaks. In order to obtain more accurate sets of data, I will fine tune the frame by adjusting each screw to its optimal roitation using a torque wrench. I will do some research online to find this optimal torque for the the type of screws used on the frame. Afterwards, the plan is to collect more data using the same procedures and compare the peak frequency values of the "better frame" with the ones I have collected here and those of the simulation.

I made a Solidworks drawing of the hanging pendulum model Kate and I discussed. To simplify the model, we didn't include the inverted pendulum. I specified the dimensions of the brass mass so that the brass mass and laser board pair had a 5kg mass at the bottom and top (each). I cut holes on the brass masses and laser boards so that the two 1mm diameter wires could be hung from above. As for specifying the materials, the downloaded laser board and struts didn't come with materials selected. So I chose "Aluminum Alloy - 1060 Alloy" for all boards, struts, and brackets. I looked up stainless steel wires on McMaster and one was "AISI 304 Steel". So I chose that for the wires. And for the masses, I chose "Brass."

Figure 1. Hanging pendulum. Origin is at the center of the middle laser board with x-axis running parallel to the shorter length, y-axis vertical, and z-axis longer length.

Under MassProperties, I found some useful properties:

Mass = 13841.72732 grams

Center of mass: ( centimeters )
    X = 0.00000
    Y = 0.01629
    Z = 0.00000

Moments of inertia: ( grams *  square centimeters )
Taken at the center of mass and aligned with the output coordinate system.
    Lxx = 14327802.07985    Lxy = 0.00002    Lxz = -0.00000
    Lyx = 0.00002    Lyy = 1189278.97915    Lyz = 0.00000
    Lzx = -0.00000    Lzy = 0.00000    Lzz = 14051480.23033

Moments of inertia: ( grams *  square centimeters )
Taken at the output coordinate system.
    Ixx = 14327805.75446    Ixy = 0.00002    Ixz = -0.00000
    Iyx = 0.00002    Iyy = 1189278.97915    Iyz = 0.00000
    Izx = -0.00000    Izy = 0.00000    Izz = 14051483.90493

Then, I used ANSYS to find the modal frequencies. The procedure I used is:
1. Select "Modal" for type of simulation
2. Import Geometry > Select the pendulum Solidworks file
3. Under "Model">"Geometry"> Change materials from library (Note: You cannot import preset materials from Solidworks. Default materials library in ANSYS is more limited. I used the pre-existing Aluminum Alloy and Brass but had to make new material called "AISI 304 Steel" and copy its mechanical properties (e.g. density, Young's modulus) from Solidworks)
4. Fixed Support > Select the circular top surfaces of the wires (the wires are fixed in space and cannot move)
5. Solve (took around 1hr)

The first modal frequency was found to be 38.44mHz and the pendulum swung in the direction of a playground swing swinging back and forth. The second frequency was 1.5063Hz in the direction of a playground swing swinging side to side.

I put together part of the rhomboid (the hanging pendulum) and added the top strut to the cage (Fig 1).

Figure 1. The rhomboid (without top board) and cage.

I've also atttached the ANSYS modal frequency simulation videos of the rhomboid with the top of the wires constrained in space (turn on repeat to play video continuously). The frequencies are as follows:

Mode 1: 38.44 mHz

Mode 2: 1.51 Hz

Mode 3: 10.15 Hz

Mode 4: 11.62 Hz

Mode 5: 12.69 Hz

Mode 6: 12.69 Hz

I've also attached the eDrawing file of the rhomboid.

We successfully suspended the rhomboid this evening. We are using only 1 wire for now and made it very bottom heavy to avoid any tipping over problems. (This was also necessary because the top breadboard does not have a hole in the middle for one wire--the locations of its holes are for the 2 wire design.) We got some good experience learning some of the limits of the pin vises, and uncovered some problems we'll face when using 2 wires. 

The first attempt to hang the rhomboid failed. The wire slipped right out of the pin vise at the bottom suspension point. This was the pin vise we had first put the wire into and we had not tightened it as much as the top pin vise. This was tightened essentially by hand. For the top pin vise, we used a pair of pliers, putting protective material between the pliers and the pin vise, to tighten the grip on the wire. The reassembly of the bottom pin vise mount was not easy. The pin vise handle did not slip into its hole easily, which was different than before. The inside of the hole appears scratched, and we cleaned it, but there was not obvious debris creating blockage. With a pair of pliers, we could force the pin vise into the hole, but this should probably be made a bit larger in the future. 

For the second (successful) attempt, we first let the bottom pin vise clamp hang freely from the wire. This allowed the wire to untwist. Its final resting place was about 45 deg from normal to the cage. We had the rhomboid supported by some boxes underneath so the topology was correct, but the rhomboid was out of the way. We then screwed the bottom clamp to the rhomboid, supported the rhomboid with our own hands, removed the boxes, and slowly lowered it until the wire supported its weight. As expected, it rotated into the 45 deg position and remained there. We have a pile of foam underneath, so that the rhomboid will fall less than an inch and onto something soft should the wire break. 

The untwisting of the wire made it clear that this aspect will pose a challenge for when we use two wires. Both wires will have to be untwisted in the same particular way so that the rhomboid sits squarely in the cage and so that neither wire is twisted in the resting position. I'm not sure yet how we'll address this problem.

We've left the rhomboid suspended. We've put a large danger sign in front of the cage to alert anyone who might enter the lab that wire is under tension and laser safety goggles must be worn.

Pictures follow:

Fig 1. Pin vise clamping onto the wire

Fig 2. Rhomboid hanging by a wire

(More pictures on ligo.wbridge@gmail.com > https://drive.google.com/drive/photos)

Kate and I suspended the rhomboid with two wires [see photo 4]. The general procedure was:

• From the previous hanging, one wire was already set up (clamped on both ends)
• Secure second wire into top pin vise. Tighten pin vise. Secure vise into holder using nuts
• Measure length of the first wire (already set up)
• Cut and secure the other end of second wire into bottom pin vise. Tighten pin vise
• Screw in the bottom vise holder onto underside of the middle breadboard
• Insert bottom pin vise into this holder
• Untwist the two wires
• Screw in the top two vise holders onto top bar of cage
• Slowly lower the rhomboid to suspend

Some key aspects include:

• Securing top pin vise into its holder using nuts
  • The relative position of top pin vise and its holder was determined by first tightening the bottom nut onto the vise. The distance between the top of the pin vise and the nut was measured and made same for the two pin vises. After slipping in the pin vise into its holder, the top nut was tightened. Photo #1 shows that the two pin vises nonetheless don't seem to be the same height.
  • TO-SOLVE: What is the most consistent and accurate way to make sure the relative positions of the top pin vise and its holder are exact? 
• Securing bottom pin vise into its holder using set screws
  •  The relative position of vise and clamp is preset since vise lip should be flush with the holder counterbore. However, because the vise had a thin, chamfered lip, it was difficult to make sure the vise was perfectly perpendicular to the holder, espeically because when we tightened the set screws, the screws pushed the pin vise around in the holder.
  • TO-SOLVE: How to make sure bottom pin vise is square with its holder? Better method than two set screws per bottom pin vise holder?
• Same length of the two wires
  • We want the wires to be of same length. One wire was already clamped on both ends from a previous hanging. We obtained a drawstring, held it next to the clamped wire, cut the string to the same length (between the tips of the pin vises on both ends), and used it as a reference length to match the second wire to be the same length
  • TO-SOLVE: Better way of making sure the two wires are of the same length?
• Tightening the pin vises to grab the wire
  • After slipping the wire into the pin vise tip, we grabbed the pin vise with two pliers with grips, held on in place, and twisted the other to tighten. We wrapped thicker kimwipes(?) around the vise in case the pliers would damage the vise. This method tightens the vise to clamp the wire securely significantly better than tightening by hand (when tightened by hand, the wire slipped out when the rhomboid was hung). We tightened the vise until the kimwipe slipped on the vise (the maximum tightness we could acheive).
  • TO-SOLVE: In the future, we want to use an adjustable vise grip with a torque wrench attached to it so that the tighteness can be measured and remain consistent for every testing. The details of attaching the torque wrench onto the vise grip and what value of torque to apply is to be determined.
• Untwisting the two wires
  • When the rhomboid is hung, we ideally want the wires to be not twisted. This time, we attached the bottom of the wires on the rhomboid (and top wires not attached to cage yet), flipped the rhomboid upside down, added mass on the top pin vise holder so that the wire was at its center of mass, and let it untwist itself (see photos 2&3). One wire was 180 degrees opposite from its desired orientation, and the other wire was about 165 to 170 degrees opposite from the desired orientation. As a result, we decided to flip the top vise holder 180 degrees (so the two wires at the top ended up farther apart than we had wanted it to be due to the asymmetry of the holder) to have the wires least twisted.
  • We came up with a better solution for next time. We would attach the top of the wire to the cage first, and let the bottom pin vise untwist itself. When untwisted, we would secure it into its holder using the set screws.
• Other TO-SOLVE's:
  • What is the best mechanism to adjust wire length once everything is set up?
    • We predict that it will be very difficult for the two wire lengths to be exactly the same, even with measurements. So after hanging the rhomboid, we plan to use a level to see whether the rhomboid is tipped. If it is tipped slightly (meaning the length of wires are not equal), we want to adjust the wire length without deconstructing everything. Adjusting the wire length at the bottom seems easier than at the top since the bottom pin vise can easily be removed by unscrewing the set screws from its holder.

Today, we tried to hang the rhomboid with the top two wires closer together (~4cm apart). First, we unclamped the bottom pin vises and rotated the top pin vise holders by 180 degrees so that the pin vises were closer together. We then let the wires with bottom pin vises attached hang freely to their untwisted state. When untwisted, it was obvious the wires wanted to be in a certain orientation (high rotational spring constant). We marked the pin vises at their untwisted states. We modified the rhomboid itself by attaching the top bread board (previously unattached). We then clamped the bottom pin vises into the bottom holders. When we let the rhomboid hang, one wire wasvisibly shorter (~1cm or so). Without letting go completely, we unclamped and clamped the longer wire pin vise, making sure to hold down the pin vise to its lowest point. When we hung the rhomboid the second time, the wires were more similar in length. Nevertheless, there was an obvious tilt to the rhomboid. We will try to balance the rhomboid tomorrow.

We also attached corner brackets to the top of the cage's four middle struts.

As we tried to balance the rhomboid, we found out that the difference in the wire lengths was the main contributor to the tipped rhomboid. Thus, we will now focus on making the wires have the same length by setting a wire clamp above the top pin vises. The idea is to extend each wire through the pin vise and top strut and hold it with a clamp made up of two plates. We can fine tune the effective length of the wire by sliding the plates up or down. When the optimal positions of the wires are found, we can then permanently secure the wires by tightening the top pin vises. The rough set up for this mechanism for one wire is shown in Figure 1.

Figure 1. We can adjust the height of the clamp using the micrometer to fine tune the effective wire length. In these photos, the clamp is incomplete; the existing platform will serve as one plate and we will need to machine the other plate. We will then attach these two using screws in order for the assembly to act as a clamp.

TO-DO:

1. Drill two holes on the top strut so that the wires can pass

2. Make a steel plate to attach to the platform (see attached CAD Drawing Wire Plate.PDF)

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

We also decided to order 6" diagonal braces to act as inner corner brackets for two opposing sides of the cage.

We will also try to attach a strut on the top side of the cage and attach 90 degree plates, on which we will attach rubber to act as stoppers.

(Updated CAD Assembly of cage + rhomboid + thermal shield to be uploaded this weekend)

I updated the cage and rhomboid CAD assembly with the latest changes. I also made a CAD of the thermal enclosure prototype. It includes aluminum panels on six sides (0/08in thickness), foam on top of the panels (1in thickness), and feet under the cage. I made some holes on the aluminum panels so that we can fasten the panels to the cage using screws. I also added corner brackets to connect the panels.

I had trouble making 3D renderings, so I took some screenshots. I may try to attempt 3D rendering again later. I've attached some photos and eDrawing files.

Updated assembly of cage with suspended rhomboid.

Thermal enclosure. Blue are aluminum panels, brown are foam.

This post will host plots and trends from this radiative cooling run (QIL/2704).

Preliminarily, it looks like the reconfiguration to remove a hardware mistake or two led to a healthier run. The comparison below clarifies the two runs:

• QIL/2702 - conductive link between inner shield and outer shield (twisted pair from an RTD lead accidentally clamped); possibly another conductive link between outer shield and baseplate (outer shield more wobbly than usual on spacers)
  • this data set should only be used to study the impact of a known conductive link between inner and outer shields.
  • this run demonstrates that there will be more effective, faster cooling if the outer shield is conductively cooled!
• QIL/2704 - resolved above mistakes!
  • this data set may be used to gain understanding of the impact of emissivity changes to the inner surface of the inner shield.
  • may be compared to QIL/2695, a run that is equivalent except with a higher emissivity inner surface of the inner shield

Run ended with cryocooler shutdown at 12:27 pm (actual duration just under 92 hours). System will warm up with pumps on for the rest of the break, unless I am inspired to come in and run one of the next intended runs discussed in QIL/2704. I did not run any heat input test for this data set, as I am not planning to come in frequently enough to monitor the heating safely.

Data:

Attachment 1 compares QIL/2704 (solid) to QIL/2702 (dashed). As expected, the outer shield temperature from the latter run stays warm since the conductive short was resolved. Due to the reduction of the inner shield's thermal load, the inner shield is able to cool faster and plateau at a colder temperature. As Stephen pointed out, however, the test mass is not cooled as efficiently compared to when the outer shield was conductively cooled.

Fitting Results:

Attachment 2 is a current model diagram of the various components being considered, and their thermal couplings. Attachment 3 plots the fitted model (dashed) over the temperature data (solid). The fit parameters were the following emissivities: aluminum foil, rough aluminum, and aquadag. Notes from the fit:

1. With the conductive shorting of the outer shield resolved, the model (which considers only radiative cooling of the OS) is well fit to the OS temperature data. 

2. The inner shield model is missing some key term(s) affecting its time constant and steady state temperature.

3. The above error propagates to the test mass model (I believe). 

Given these caveats, the fit results are as follows: aquadag e = 0.92, Al foil e = 0.04, rough Al e = 0.19. These all initially seem reasonable, and I'm happy to see that the aquadag emissivity is higher than previously estimated.

Next steps:

1. Separate the cold plate from the inner shield, and model their conductive and radiative link. Also model the radiative link between the cold plate and the test mass.

2. Cover the test mass in foil (to best of our ability) to refine the radiative link between the test mass and inner shield. Doing so will mean both elements have the same emissivity, so there is only one unknown parameter.

Torque driver set for QIL setup bolted joints, with range 15 in*oz - 50 in*lb, p/n WIHA 5HYL9, is on order from Grainger, with anticipated delivery in the week of July 20th. Refer to  PO S477925. *update* Tracking Number UPS 1Z19W9330321365493

Cryo connection copper parts PO S475316 will be finished early next week by the machine shop in Torrance, I'll bring them to campus or to Raymond's place (TBD).

Copper parts picked up July 23rd and brought to QIL, now only waiting on PO# S477874 and the pirani gauge from Koji's bulk JPL order