At some stage we will need to install new temperature sensors on the thermal shields.  Tara et al. had installed AD590s in vacuum (and did a similar thing for the pervious longer cavities), these are now broken. They give currents proportional to applied voltage in the forward direction (see: PSL:1173). One theory is that they were overheated during soldering. Another possibility is that that the ICs didn't fare well in vacuum.  Either way it would be best to have passive components that we know are vacuum compatible and low noise.  

Previous rana post here on temperature sensor noise plots: PSL:1205

We need a robust low noise circuit to turn resistance into a ADC usable voltage.  I've attached a sketch below. It uses a IC dual current source regulator (REG200) and a low noise instrument amplifier (AD620) to convert resistance of a PT100 sensor (like the one used in the TCS, see PSL:1700 and subsequent posts) in a three wire scheme. I've yet to compute a noise budget, for now I've penciled in some specs for noise I've drawing from their data sheets. This scheme would require three wires per sensor (but it might be possible to double up the grounding line between sensors). 

Note: the R1 resistor provides an offset so that the Vout is within a useable range for high gain. R_g sets gain, where G = 49.4kΩ/R_g +1.

edit: added not on function of R1 and R_g - awadeWed Nov 9 17:47:14 2016

It appears the acromag1 box was reset again.  This might have to do with the flooding in the adjacent ATF lab stopping power there.  

I have ordered the APC SMX3000LVNC UPS unit.  The UPS tracking says it arrived Wednesday last week. My defults for TechMart punchout were missing the LIGO mail code so stuff is taking a bit longer to get to me through the internal mail system.  Should be here this week.

The APC SMX3000LVNC UPS unit arrived today.  It will need to be tagged and put into the LIGO asset system.

The question is where to locate in lab. Unit can be mounted vertically like a tower computer or rack mounted (obviously at the bottom as it is ~42 kg).  The lasers appear to be powered off a strip down the side of the table. The rest of the essential stuff is in the rack so that would be the place to put it.  There is already a bunch of heavy duty HV power supplies at the bottom of the rack, they will either have to be moved (not a great option) or the UPS might be installed above.  

Not sure how much more it will weigh when I charge it.

I picked up one of the following linear lab power supplies for the TCS Lab. They can output 30V, 3A. They also have accept an analog input to control the output voltage (or current). It's the sort of thing that should be useful for quick heater driver tests.

http://www.newark.com/gw-instek/gps-3030d/power-supply-dc-30v-90w/dp/73K6023

If we wanted to measure the noise of this without saturating the network analyzer I'm guessing we put a capacitor in series to block DC. What kind + size of capacitor should one use? When I dump some numbers into partsim 1 uF (with 1 MΩ load) gave a cutoff of 0.125 Hz, does that sound right?

Yinzi is going to make some mesurements of the Acromag card noise on Monday.  Maybe we could do that at the some time...

The (Windows 7) computer that runs the OPC server is in the TCS Lab. The EPICS server on this machine needs rebooting at the moment. 

More critically, we need a framebuilder running on the network in order to save these channels to file for future trending. A slow EPICS framebuilder is all that is necessary.

I've pulled the old PMC driver board from the VIM crate architecture and inserted it into a chassis (D1700003). The interface has been reconfigured to insert directly into two Acromag units that are inside the chassis (see attached photo). It still needs a back panel.

So far, the DAC connections (loop gain control and locking point control) are wokring. The binary outputs are not properly switching between 0V and 5V - I'll need to double-check what circuit the Acromag 1541 is expecting to be attached.

Also, I've added a new EPICS database file for these channels (attached).

Attached schematic for photodiode PCB labeled "Rev. 0 LIGO Laboratory LSC Photodiode PCB 26-Mar-1998"

This is sort of complicated to debug. Because of the postive feedback on the bias line it can be tricky to debug and tune. IF the positive feedback tuning is wrong then the resonance of the RFPD can move once you put light on the diode.

This is an overview of the old noise budget made by Evan and Tara.  The plot is the result of noisebudgetQWL.ipynb written by Evan.  QWL stands for quarter wavelength, referring to the coating layers' thickness (see Fig 2 of Evan's paper).

Each curve is given an brief statement about it's origin.  Here is a link to Tara's paper on the PSL Lab setup.  Here is a link to Evan's paper.  These papers have convenient tables with parameter values for the setup and reference cavities. 

The x axis is frequency in Hz, the y axis is the ASD in Hz / rtHz, aka frequency noise.  Many of the thermal noises are reported as length noise.  To convert from length to frequency noise, use Delta f / Delta L = c / (L * lambda). 

COATINGS THERMAL NOISE BUDGET

TOTAL EXPECTED (blue):  Sum of all expected noises.  This does not equal the actual measurement in red, meaning not all sources of noise are accounted for in this plot.  One suspicious missing noise source is scattering.  Not much has been done to mitigate scattering in the PSL Lab setup.

MEASUREMENT (red): Actual beatnote measurement measured using a phase locked loop with the cavities' transmission radio-frequency photodetector (See Fig 2).  The two lasers are locked to their respective cavities to reduce the free-running laser noise via PDH control loop gain.  By suppressing laser noise, we can reveal the residual cavity length noise, hopefully dominated by broadband thermal noise.

COATING BROWNIAN (green): Theory curve of the estimated coating brownian noise.  Brownian noise magnitude is governed by a material's mechanical loss.  "Mechanical loss" refers to rate at which kinetic energy in a material is "lost" to thermal energy.  Unclear why Equation 8 in Tara's paper and Equation 3 in Evan's paper are different, I think it has to do with assumptions about the coating and substrate Young's modulus and Poisson ratio being the same.  In the noisebudgetQWL.ipynb, Tara's coating brownian noise equation is used.

COATING THERMO-OPTIC (pink): Theory curve of the estimated thermo-optic noise.  Thermo-optic noise comes from temperature fluctuations in a material causing cavity length changes.  This seems to be the key curve to all Tara and Evan's work.  The idea here, originated by Evans et al., seems to be that thermoelastic noise and thermorefractive noise can cancel one another in thin enough coatings.  Given by Equation 9 in Tara's paper and Equation 4 in Evan's. 

SUBSTRATE BROWNIAN (yellow): Theory curve of the estimated substrate brownian noise.  Like the coating brownian, but refers to noise originating from the fused silica making up most of the mirror.  Equation 5 in Tara's paper.

SUBSTRATE THERMOELASTIC (teal): Theory curve of the estimated substrate thermoelastic noise.  Thermoelastic noise refers to how temperature fluctuations cause a material to modulate its length.  Governed by the coefficient of thermoelasticity alpha = 1/L(dL/dT)  Equation 6 in Tara's paper. 

Incidentally, I will mention THERMOREFRACTIVE noise here since there is no curve dedicated directly to it, but it is important to thermo-optic noise.  Thermorefractive noise comes from temperature fluctuations changing the refractive index n of a material light is passing through.  Governed by the coefficient of thermorefractivity beta = dn/dT. 

POUND DREVER HALL SHOT NOISE (orange): Theory curve of shot noise.  Shot noise refers to the Poisson statistics of fluctuations in the number of photons incident on a photodetector.  This noise PSD is flat in frequency, but falls as 1/f^2 in power.  To convert from the power PSD to frequency PSD, multiply the power PSD by (1 + f^2/fc^2)/(2 P0 Gamma/fc)^2 where fc is the cavity pole, Gamma is the modulation depth, and P0 is the incident power on the cavity. Equation 20 of Tara's paper.

PHASE LOCKED LOOP OSCILLATOR NOISE (grey):  Measured noise from the PLL, presumably originating from the voltage-controlled oscillator (VCO).  Figure 5 in Tara's paper shows the PLL and the various noises found in it, including photocurrent shot noise, photodiode amplifier noise, and VCO frequency noise.  Unclear what the 707 Hz/V means, probably is the VCO control slope (i.e. if I want to change my VCO freq 707 Hz, I raise the control voltage by 1 V).

PHASE LOCKED LOOP READOUT (purple): Theory curve of the PLL readout noise.  The PSD for this noise rises as f^2, due to the fact that the PLL is a phase detector but the noise budget is in units of Hz/rtHz.  This curve is poorly documented compared to the rest of them (Evan calls it a "magic number" curve).  To convert from phase noise to frequency noise, multiply the phase PSD by f^2.

SEISMIC COUPLING (black): Measured curve of the seismic coupling into the experiment.  The raw data taken appears to be seismic velocity in units of m / (s * rtHz) as a function of frequency.  Then, seismic acceleration is obtained by multiplying the raw seismic velocity data by 2*pi*f.  Then the two stacks (?) and a spring (??) TF are modeled with hard-coded resonant frequencies and Q's and multiplied together to give a final seismic TF falling as 1/f^6.  The final seismic PSD is found by squaring the product of seismic acceleration, the 1/f^6 seismic TF, and an additional hard-coded seismic coupling factor dependent on the cavity length with units  m/(m s**-2).

PHOTOTHERMAL NOISE, ISS ON (brown):  Measured curve of the photothermal noise.  Photothermal noise originates from fluctuations in laser intensity causing changes in the amount of laser power absorbed by the coatings, which causes coating temperature fluctuations.  Seems to be the expected limiting noise at low frequency.  The raw measurement was of relative intensity noise from both lasers.  To get the photothermal noise PSDs for each path, the RINs from each path are multiplied by the absolute laser power absorbed by the coatings squared, a "total photothermal TF" squared, and converted from length PSDs in units of m^2/Hz to frequency PSDs Hz^2/Hz.  The "total photothermal TF" is the sum of the coatings photothermal thermoelastic TF, coatings photothermal thermorefractive TF, and subtrate photothermal thermoelastic TF.  Each of these photothermalTFs are from units of transmitted cavity power to beatnote frequency fluctuations, (i.e. Hz/W).  The process of measuring these three TFs is explained in Evan's paper, section VI, subsection A.  It seems that the successful cancellation of expected photothermal noise was the main success of Evan's paper.

RESIDUAL NPRO NOISE (dark green): Theory curve for the residual NonPlanar Ring Oscillator (NPRO) laser noise.  The freerunning ASD of the NPRO is reported to by 10**4/f with units of Hz/rtHz.  This is then divided by the PDH control loop gain for both paths, squared into a PSD, and summed together into a final NPRO residual PSD. 

The coating thermo-optic trace seems off here. The idea is that the actual coatings we have installed are the ones with the cancellation as detailed in the thermo-optic cancellation paper.

It seems like the QWL noise budget you're using is the one that shows how bad it would be if we used quarter wave layers, so you ought to email Evan and get the latest one which has the as-built coatings.

I built a temperature sensor prototype for the 40m lab, which can be used for the PSL lab as well for temperature stabilization. It consists of an AD586 5V constant voltage output, an AD590 temperature sensor (I initially had 592 but they are very similar), and a LT1012 OP amp, along with a 10K resistor and a few capacitors (see first schematic). It uses the temperature of an object that touches the AD590, which is attached to a long cable, and converts it into a current (1microA/K), which is converted by the OP amp and resistor into a voltage. The AD586 is required because the sensor wants a constant input of 5V to accurately measure temperature. I used a 10K resistor on the OP amp so that the voltage should be around 3V when measuring room temperature. It requires an input of +15V and -15V to the OP amp, 15V to the input of the AD586 (this is shared with the +15V input to the OP amp), and a ground. The +15V (red) and -15V (black) inputs are given to the pins closer to LT1012 and the ground is farther away. The output voltage is read out through a BNC cable and can be converted into a temperature in K by multiplying the value by 100, but it shows a temerature that is about 2K higher than the actual temperature.

In addition, I used capacitors on the OP amp to stabilize the voltage input. I used a 100nF ceramic capacitor placed close to the pins of the OP amp and a 100microF electrolytic capacitor placed father away to achieve this. I have attached the schematic for this as well (second schematic), with 1 being the electrolytic capacitor and 2 being the ceramic capacitor.

I've put together a summary block diagram of the FSS boxes (based on the schematics of our version). This summarizes all the stages of filter and amplification as they were in the original design.  There have been modifications since then which we are working to document.  Craig and I will look at the boards again tomorrow and stick a probe in to measure at various points to confirm.  Craig and I have also been looking at LISO files (and porting this to a python wrapper) and he has also almost got a full set of plotted liso TF for each board and each stage on each board: this will help us breakdown what is going on.  He will post soon.

We also need to follow through with improving the cross over characteristics of the loops. It looks like we should reexamine the choice of components for the boost that was installed (by Frank et. al) at U7 on the PZT path. Its not clear to me why they chose that op amp to apply boost, given that it is already handling a notch there. I guess its fine where it is, but the values should be looked at. The values of resistors and caps in the boost are also not the same between north and south path.  Maybe we could also look at lowering the second stage low pass (U9) in that path from 34 kHz down to maybe 15-18 kHz.   Evans, in his squeezer FSS modification, added a mini-boost to U8, which seems like a good place to do it: unless there is something like saturation and railing in the later stages of amplification there. I think they had their main boost on the RF board, which is a different make to ours; we have just one summing op amp in that path.

Then, in the EOM path we can increase the AC coupling (twice) by lowering C23 and C24 while raising the first stage pole up to a much higher frequency by lowering R19. Evans (post) did this by lowering R19  (also lowers gain), not sure why can't lower c15 but its already at 3.3 nF.

Another thing I don't understand is why the notch is so high in the PZT path. I would have thoughts the resonances and peaks in the laser PZT response would be much lower, on the order 10s of kHz.

Edit (awade, Mon Sep 18 13:39:59 2017): These are modifications to the North path field box

Craig has been working on active probing the FSS boxes to work out what is going on there.  He came up against some unexplainably low gains and stuff that didn't quite match up with what we should expect when measuring transfer functions between TEST2 EXC (the common path excitation point) and TP5 and later test points in the box.  

A while ago I had found that measuring and tuning out1/out2 OLG with the Agilent plugged in would get to a stable loop, but as soon as I unplugged the Agilent it would be unstable.  I assumed that the monitor points needed to be 50 Ω loaded when not being measured: which is what I did as a default.  

The open loop gain from common EXC to TP5 on the servo board was -3.15 dB.  With common OUT1 terminated with 50 Ω this TP5 dropped to -17.8 dB or with OUT2 it dropped to -25.75 dB. Bad. This was a clue that maybe the series resistance to the monitor outputs was selected too low.

Now looking at the changes that Frank made to the schematic (PSL:893), he had changed the series resistors R1 and R2 from 453 Ω to 50 Ω saying that the monitors were limited by thermal noise.  I don't know if I believe that. We should have plenty signal to noise. Also, with 50 Ω to monitors a 50 Ω load of the network analyzer will have 100 Ω to ground. I have a feeling that this is causing the op amp to be over drawn in current. The AD829 datasheet says max current 32 mA.  I don't know if the servo signal at the point gets to 3.2 V offset but it might. It seems like it would be better to have higher load impedance to avoid overdrawing current out of the op amp output. 

I switched the R1 and R2 50 Ω (on D040105-C) out for something higher again, this time 470 Ω.  I could only find 470 Ω in the EE workshop (which is close enough to the design value of 453 Ω).

Now plugging 50 Ω loads into the OUT1 and OUT1 common monitor points lowers the common path OLG by 7.8 dB and 10.8 dB respectively.  This still seems like a big change to me but I'm not sure if I should increase series resistance any higher.

Let me know if this is dumb and I'll change it back.

The noise budget posted in the previous noisebudget elog has been described by experts as "bunk".  I am here to right the ship with a more correct noise budget.  I have committed the corrected noisebudget code in the CTN_noisebudget git under /CTN_noisebudget/noisebudget/noisebudget.ipynb.

In particular, the pink curve in the new noise budget is the real COATING THERMO-OPTIC NOISE.  It is far lower than before.  This is why the optimized coating layers were created in the first place: to strategically cancel coating thermoelastic and coating thermorefractive noise.  Recall that thermo-optic = thermoelastic + thermorefractive, and they can cancel if their phases are carefully controlled.

With lowered COATING THERMO-OPTIC NOISE, we can now think about measuring COATING BROWNIAN NOISE for the AlGaAs coatings. 

I have included a new trace: the SEPT 2017 BEAT MEAS.  These are just some preliminary PLL noise limited spectra to keep you excited for great low noise measurements to come.

***

I spent a bunch of time making an interactive noise budget bokeh plot.  HERE IT IS. 

Advantages: You can scroll over and see exact numbers at whatever frequency you want.  You can pan and zoom.  You can click the legend to remove lines you don't want to see.  It looks nice.  We can host these on our 40m server soon (I'm hosting this on the ldas cluster, so you have to login).

Disadvantages: File is huge (~20 Mb) and takes a long time to load.  Interactivity is choppy.  Panning causes lines to disappear.  Tick labels are cancerous.  The dark background is good only if I can figure out how to make the entire background dark, this means making custom CSS files and I'm no html expert.

[Aidan, Gabriele, Eric]

We turned off and covered the clean flow bench by the west wall. We also removed the items from off the top on the half closest to that wall. We removed all the cables hanging below the pipe and lay them on the floor.

good, but please no LDAS links in this elog; we don't want to be tied down to ligo.org and their mercurial ways

for things which require permanence, you can just make a directory in the elog directory on nodus and put your Bokeh stuff in there

The text wrapping issues is back, PSL:1919 post had a line of *** unbroken without spaces that forced a super wide page.  Use the horizontal line button next to the smiley faces to make breaks in text.

PSL:1745, shifts returns were removed as well as zipping the attachment

PSL:1644, text only attachement was zipped and reuploaded

PSL:1641, text only attachement was zipped and reuploaded

PSL:1100, shrink-rayed ludicrously large image

PSL:160, fixed shift return, text copy-pasted from non-text wrap source

The purge is complete. 

Ugh - no, plugging in the network analyzer should not change the gain of the servo. This seems quite bad. Lets discuss and fix.

In addition to Frank's fixes, Tara also did some patches with advice from Koji. We should make sure you have the right version of schematic.

In any case, once you have R1 and R2 at 450-500 Ohms, the circuit should not change behavior just by adding 50 Ohms onto EXC or OUT2. You should be able to do the TF without OUT1.

Yes, this is bad. I thought maybe it was an issue but didn't address it as was working off the original design schematics. Also, originally I was using the SR785 to start with.

Information on all the changes made is spread thinly over the elog. It hard to search as terms used and catagories change over time: I just need to do a more detailed search. I've made a ATF wiki summary page HERE to bring together and summerize what we do know. Future changes should be document cronologically and in detail on the elog and ALSO as a live summary on the wiki.  I will add a section at the bottom of that page for a change-log where description, date and elog entries can be linked. 

When using LISO to simulate your circuits, you may want to plot a few different cases on top of each other. For this purpose, there is the perl script 'mfil' in the LISO GIT repo:

https://git.ligo.org/40m/LISO/tree/master/filter

Although eventually someone with some free time may write a python version, for now it might be best if we could just use it as is or potentially make a modified pyKAT to run it.

I had another look at the common path gain stage of the FSS field boxes (LIGO-D040105).  In modifications to the two field boxes in the PSL lab a switch with capacitor in series had been added to U7 for intergrator boost.  A gain killing switch, with 10 Ω in series, had also been added to U3 to manually kill gain while engauging the intergrator.  We suspect this -- along with only 100 Ω to ground on the out2 test point -- may have degrated the U3 AD829 chips. 

A transfer function from TP2 to TP5 of the north servo board (LIGO-D040105) was taken.  An excitation was put into the Test2 SMA (on RF board LIGO-D0901894) and a pair of HP 41800a active probes were used to take the TF accross the test points. Also, for comparison, I did the same for the FSS field box procured from the 40m (2010:006).  The data are plotted below: Ser5 is the north path serial numbered #5 board and Ser6 is the #6 board of the 40m unit.  Between the north path (2010:005) and 40m (2010:006) boxes there is a overall gain difference (due to changes in the feedback resistor) but the slope is very different.

I then replaced U3 (an AD829) on the servo board (LIGO-D040105) and retook the transfer function (also plotted below). With a new op amp the slopes of the magnitude and phase of the bode plots now match between (presumably working) 40m box and the north path FSS box. This is good.

One thing I don't understand is why there is a roll off starting at or below 1 kHz.  I used a 300 Hz resolution BW on the TF so below 600 Hz is bad.  But I've probably done something stupid in the way I've gone about this measurment.  I wanted something quick and easly repeatable, I'll probably realize tommorrow what I've done wrong.  Craig measured this about two weeks ago and it looks much flatter. What he sent me is attached below, it includes a LISO model. From the info on the AD829 datasheet we expected it to be flatter out to 1 MHz for a gain of 4.

So these seems replacing U3 on LIGO-D040105 board is a slight win but we (really just me) still don't understand why it has a sloped TF below 1 MHz.

---

I also took the opportunity to replace the R1 and R2 resistors with their design value of 453 Ω. I had used 470 Ω in a previous rework of the board (PSL:1918), but these were the wrong pitch and I did a messy job of it.  I found these 453 Ω taped to the inside of the north box.

[awade, Craig]

Yesterday Craig measured the transfer function of the south path servo board (LIGO-D040105 2010:007) at all the available test points from the input to the board. This was done with an injected signal of 0 dBm (=2*sqrt(50 Ω * 0.0001 W) = 0.45V, there is no 50 Ω loading here on input) over the range 1 kHz to 100 MHz.  The test points were probed with the hp 41800a probe. 

I've overlaid these results with expected LISO simulation; a scanned schematic with labels of stages and nodes is attached below. There is an adjustment of 3 dB made for the two port RF splitter used to spit the output from the HP4395A.  Also the knobs for common and fast gain stages were dialed to 500 out of 1000 in both cases. It appears that this gives pretty much unity gain for both AD602 variable gain chips.

There seems to be fairly good agreement in the common path.  In the PZT path model and measurement disagree below 60 kHz, getting to factors of 100 off at later stages of the loop at low frequencies. This is an open loop measurement, not sure if poor signal to noise is the issue at low frequencies here. Not clear what the discrepancy is here. At high frequencies we expect the PZT path to drop off, first stage op27 has GBW of 8 MHz and is configured as gain 8.7 so should only work up to UGF ~1MHz: the HF drop off in signal to noise makes sense, I guess. 

One big discrepancy is the the fact that the PZT path 239 kHz notch matches well at the test points, but the full test2 to PZT out BNC sees the notch shifted downwards.  This is suspicious. 

The EOM path agrees pretty well along the main path except for poll at 40 kHz. We should check again on the choice of rc components incase we missed a design change. There are no test points on the passive loop over path, we should maybe make a measurement at node 10 or 14 to see what the passive filter elements are doing.

---

I had a look at mfil, looks like it achieves the same as our python scripting but requires customizations of the fil files. Now with Craig's parsing methods in pyliso we can arbitrarily modify any component  from python wilst leaving the base FSS_all.fil liso file alone. I'll eventually get around to testing/intergrating mfil against pyliso, but for now they are about the same amount of work to use: at least pyliso can imediatly be used for pretty plots.

---

Data attached as zip. You can also clone current repos from https://git.ligo.org/andrew-wade/CTN_noisebudget and https://git.ligo.org/andrew-wade/ctn_labdata

probably saturating in the PZT path. Repeat with -10 dBm source and look at outputs with a scope to check if its internally saturating.

Yes, saturation was the issue.

I turned down the input excitation to -23 dBm and checked with test clips all the way through the the test points on both paths.  Looking with the oscilloscope The only point saturating was TP13, the last stage of the EOM path.  For this data point I turned the power down to -50 dBm; I have a feeling this is too low and maybe we need to injected an excitation at at the start of the EOM to get a decent TF.

I've attached TF for both north and south field boxes. 

South box (2009:007):

There is still a little saturation on stage 3 of the PZT path, it seemed fine looking with the oscilloscope so I'm not sure what is going on there: its possible I was looking with a 100 kHz excitation and that it looked ok, but was not at lower frequencies.  Something to check again. It also looks like EOM stage 2 is saturating.  On the third stage EOM, I turned the excitation right down (as mentioned above).  This looks not right, but might need more excitation closer to that point in the circuit. 

North path (2009:005):

As you can see, there is a fault at the first stage of the common path.  Craig has since diagnosed this to be a fault with the common path circuit.  The pad has peeled off for R4 on the pin2 side of the op amp. Craig has more detailed pictures of his before and after fix of this. Some time in the past R4 had been removed from both of the FSS boxes in the PSL lab and replaced with with a regular metal film through hole resistor pictured below. The resistor sits in vertical pin holes soldered to the pads. I guess the motivation was to have some programmable gain, but the resistors has been knocked and now we have some damage on the north (2009:005 unit).

---

All data and plots are committed into ctn_labdata repo in /data/20171020_Ser7_South_FSSTFs and /data/20171020_Ser5_North_FSSTFs

​Plotting notebooks are commited into ctn_noisebudget repo under /TTFSS_lisomodel

[awade, Craig]

Craig retook the transfer functions of both north and south boards today with 7 mVrms (-30 dBm) input signal into TP1. I've plotted and they are attached below. 

There is a sign flip in phase in the north path stage 5 due to the flipper switch being in the opposite state.  Things seem to aggree except for the EOM final stage 3.  We need to check the liso model here and also the actual components on the board.  As noted by Craig in PSL:1956, some of the capacitors in the zeros of the main op amp stages are higher value than in the schematic (twice as much).  This pushes the frequency down, we actually probably want to go in the oposite direction. We also can't see the EOM path notch in the liso model. That is a red flag.

Fil file and plots attached. The rest is commited into the ctn_noisebudget and ctn_labdata gitlabs.

I had a guy come and fix the lighting in the lab.  I tried to replace the tubes yesterday and found that the ballast had died for one row of lights.

A facilities guy came this morning and replaced the ballast on the outer row on the north side of the lab.  I turned the lasers off from 10 am till 11:30 am Wed Jan 10 to make  it easier for him.  The lasers are now back on and cavities are locked. I didn't relock the PLL. 

Updated todo list is up on the labdata git ctn_labdata/issue/10.

Priority over the next week is mittigating ghost beams within the vacuum can. Its liklely we will need to vent to reangle cavities. 

Hi, I'm an undergrad and I'll be helping to upgrade the lab data acquisition system. I'm starting off with getting data from fb4, making plots of lab temperature, laser power etc which would lead to posting them into html pages.

Link to Github repo: https://github.com/shufay/LIGO-plots

- Shu Fay

quickplot.py makes quick plots of data from desired channels. See: https://github.com/shufay/LIGO-plots.

On ws1 cd to ~/Git/LIGO-plots. In Ipython:  %run quickplot.py <channel 1> <channel 2> ... <(optional) gpsLength> < (optional) gpsStop>

To see usage: %run quickplot.py usage

Arguments:

<channel 1> <channel 2> ... Channels that you want to make plots of

<gpsLength> Length of time to fetch data. Default is 3600s.

<gpsStop> GPS time to fetch data until. Default is now. So the default parameters would fetch data from (now-3600, now).

DC photocurrent:
a few mA => A transimpedance of 1k Ohm will realize the output of ~a few V.
A few mA of DC current produces the shot noise photocurrent of 20~30pA/rtHz  

RF photocurrent: 
Resonant circuit resistance @25MHz: 100~300 Ohm.
Reduced to 100 by the shunt resistance of 150 Ohm
MAX4107 required minimum gain of 10 for the stability
-> Total Transimp. R=1K
-> Andrew said R=1K gives 13dBm output (1.4Vpk @25MHz)
-> This corresponds to 200 V/us of slewrate. This is ~40% of the full scale. (Too big)

input referred current noise ~ 12pA/rtHz ==> Shot noise intercept current ~0.44mA

If we reduce the shunt resistance to get the total transimpedance of 500Ohm
=> Beat output: 7dBm (0.7Vpk@25MHz, ~20% of the full slewrate)

input referred current noise ~24pA/rtHz ==> Shot noise intervept current ~ 1.8mA

Questions:

- Can we torelate this noise level (24pA/rtHz)?

- Or do we have MAX4106 (min gain 5), or something else for a replacement?
 

MAX4106 slew rate is 275 V/us, so almost half of 4107. LMH6611 is 460 V/us.

What about THS4271-EP ? (1000 V/us, can be used with a gain = 2, Vn = 3 nV, i_n = 3 pA)

Update: The "EP" model doesn't come in SOIC-8, so I just ordered the regular THS4271. It wants a 'PowerPad' heat sink, but we can try it without and see what happens as a test.

Made the first article of the 25MHz resonant PD for CTN experiment.

Performance:

Resonant frequency f_res = 26.0MHz
Transimpedance R_res = 2.3kOhm
Input referred curent noise = 10pA/rtHz
Shot noise intercept current = 0.29mA

Remarks:

MAX4107 at G=10 gives a gain peaking at 260MHz. It is not tamed as the dark spectrum shows the peak does not have significant RMS.
But this might cause a trouble later. We can consider to replace the opamp with ones suggested by Rana. Rana bought "THS4271" to give it a try.

I've drafted up some new drawings for new view ports into our vacuum can.  We have 8 inch windows at the moment with AR coating and surface quality that could be better.  Given the criticalness of scatter we plan to replace the windows with a set of four 2 inch AR-AR coated windows; smaller windows will also reduce the exposed thermal surfaces to outside, thus improving thermal isolation of the cavity shields from the outside. 

Specifications of view port windows

Planned spec for the windows is 10-5 scratch-dig, lambda/10 flatness and AR of R<=0.25 % with a 30 arc second wedge. I've contacted a number of vendors for quotes. As we can't list prices publicly here I've made a summary page HERE with prices and specs from vendors that responded.

Design of flange and retaining ring

To deal with any residual beams from the surfaces of the window we will angle the windows at 3 degrees to normal by insetting them into the flange: centered beams on the optic that are normal the flange face will then be reflected at 3 degrees (52 mrad) which at a distances of roughly 20 cm will be 1 cm offset from main beam, this is plenty of space to fit in a beam dump. This design may present challenges for machining.  I'm pretty sure the groves can't just be milled, they need lathing with certain tool specifications to ensure there are no radially oriented scratches that might compromise the seal. The flange is a bulky piece, we need to ask a machinist what can be done.  The angled inset is the way to go, even if its a little harder to make.

I based these drawings partly off Johannes's design (see Cryo:1456). The particulars of their design meant that I couldn't reuse the parts, but I took cues from the design features (that made things much faster to draft). I rebuilt the flange parts and retaining ring with a more parametric design.  Within the part files dimensions are defined in a way that makes it easy to change wedging angles, groove dimensions and sinking depth of the optic.  Hopefully this will help future users adapt them to their needs quickly.  Johannes also had a good reference for view port designs Abbot & Scace in J. Vac. Sci. Technol. A 28, 573 (2010).  One thing that isn't addressed much in the literature is the extra loading applied on the inner o-rings when 1 atm of pressure is applied (vacuum).  For a 2 in^2 window area. 1 atm is about 20.9 kg of pressure (8.6 lb/in of oring). For 70 duro viton o-rings the additional compression looks to be on order of 7 %.  Provided there enough clearance is left between glass and metal, this shouldn't me much of an additional error: it just something to keep in mind for tolerances of the parts.

Drawings are attached below along with an ipython notebook used for the calculation of various dimensions (this can also be found in a gist on GitHub). A zip contains all the parts, drawings and assemblies.

The basic design criteria are for a 2.0" view port optic, wedge by 30 arc min, that is 9.35 mm on its thickest edge.  Clearance of 3/128" are made around the perimeter and 0.5 mm between the faces and metal of the flange and retaining ring.  The o-ring I selected is #130 which has ID of 1.612" and thickness (toroidal diameter) of 0.103".  With a set compression ratio of 0.72 for the o-ring it is possible to make an o-ring groove just deeper than half the o-ring thickness. An o-ring any smaller will be more than double any possible groove depth, which would be bad as would will pop out during assembly. Dimensions of the o-ring groove were chosen according to rules outlined in Abbot & Scace. I think at a 2" diameter it is a little too small to make dove tale o-ring grooves, so the inner diameter of the groove was made 2% larger than the o-ring so that it will be under a little tension while the unit is assembled.  The outer edge of the o-ring groove is 2 mm from the edge of the optic, this should be enough space for the o-ring to be clear of any edge imperfections. Calculations for all these quantities can be found in the ipython notebook.  

The rest of the dimensions were adjust accordingly to reach the correct clearances, an optic angle of 3 degrees to the front of the flange and leave 2 mm protruding on the lowest side of the optic.  The two windows are oriented about the center line of the 10" flange blank, spaced 3.0" apart.  The angle is currently set to point outwards on the outside  of the tank; this orientation might be wrong for our needs as we want to primarily dump light inside the tank and it would be easier to put the beam dumps not between the two beams.

I need to stop fiddling with this now.  It would be good if someone could look over the drawings and any raise issues. 

MAX4107 is stable only when G>+10. Because of this fact, the 25MHz PD has a too-high transimpedance for our purpose. To lower the gain without losing the stability of the amp, I have implemented a compensation network.

Attachment 1 shows the compensation network. This 10Ohm+100pF pair effectively makes the high freq noise gain higher (~10) while the low freq gain is G=4.5.

The actual implementation is found in Attachments 2 and 3. The overall schematic can be found in Attachment 5.
I also chose the resistor values so that their noise contributions are reduced.

The resulting output spectrum is compared with the one before the modification (Attachment 4).
Even with the lower gain, the gain peaking of the amplifier output is reduced.

LISO model (not shown here) indicates the input referred noise is ~1nV/rtHz. Considering the voltage division by the 50Ohm termination, the output voltage is as low as 2.25nV/rtHz. This is not a high number. Thus, we practically need a mid power / low noise amplifier attached to the output of the unit.

The total performance of the PD unit will be tested again shortly.

Performance of the modified photodetector unit:

Performance:

- Resonant frequency f_res = 26.0MHz
- Transimpedance gain at 26MHz is 1.27kOhm (560Ohm resistance of the resonant circuit x G=4.5 x 1/2 by 50Ohm termination)
- Input referred curent noise = 9pA/rtHz
- Shotnoise intercept current is 0.24mA

Remarks:

- There is a gain peaking at 280MHz as explained in the prev elog. If one does not like this, remove 700Ohm resistor from the max4107 stage. It will increase the amplifier gain from 4.5 to 5, and the gain peaking is reduced.

- The transimpedance gain might be still too high. Then, a shunt resistor at the location of R24 can be added to the resonant circuit. This way, the shunt resistance is seen only from the RF path. Of course, lowering the signal level has to be paid by the increase of the input referred current noise level.

Some updates for 2 inch window flanges.

Errors fixed

I noticed a few small errors in the 2 inch window flange design.  I hadn't factored the clearance of the window from the metal (0.5 mm) into the depth of the o-ring grooves, so as to get exactly 0.72 compression when assembled.  Also the clearance on the retaining ring side was wrong because I had computed the angling depth on the lower side of the optic based on the optic dimensions and not on the size of the cut hole with clearance.

To fix these issues and clarify the specified clearance, in the solidworks part itself, I have made the clearance explicit with two additional revolve cuts. One is around the face and around the circumference of the optic. This way the clearance will also be parameterize in the part: this should improve adaptability of the part to other applications. Groove depth is now set at the desired compression ratio and the correct real depth of the final groove is realized by the clearance cuts.

Also I changed the angling of the two windows to point inwards on the outside of the can.  This means that ghost beams will be maximally separated on the inside of the can, making it easier to mount a pair of beam dumps either side of each cavity.

Files uploaded to DCC

I've put the drawings and assembly on DCC for better version tracking. I've attached the solid works folder below in a zip for local reference.

• LIGO-D1800100-v1 Assembly document
• LIGO-D1800098-v1 Flange 10 in , 3 degree angled inset window
• LIGO-D1800097-v1 Clamp retaining ring, 3.5 degree angled, to go with 0.5 deg wedged optic
• LIGO-D1800099-v1 Window 2 in Viewport, 0.5 degree wedged

How it fits together

I have yet to finish checking the whole tank assembly.  For now here is a pretty animation showing port and orings.

This is a summary reference post for parameters to do with the thermal surfaces and bodies within the vacuum can. It brings together drawings and computed dimensions so we can begin to make an actuate physical model of the thermal dynamics of our system.

Design goals

At the center of the experiment are a pair of Fabry-Pérot cavities that need to be thermally stable enough to not drift more than 100 Hz in the time it takes to take a PSD of their relative brownian driven fluctuations.  

\Delta T = \frac{\lambda}{2c\alpha} \delta \nu

https://nodus.ligo.caltech.edu:8081/CTN/1874

Overview

[Insert SW cutaway with ballons]

Cavity parameters 

Cavity cylindrical heat shields

https://nodus.ligo.caltech.edu:8081/CTN/1737

I discussed the 2-in window flange design with Steve.  He had a couple of suggested changes.  I'm adding them here for reference as I update the drawings.

He says the angling of the optic into the main flange is fine, 8" 40m flanges have this design.

Changes:

• The retaining ring should be made of Delrin, this will be more forgiving on the optic when at atmosphere if there happens to be any contact points.  Also easier/cheaper to machine.  The retaining ring is only there to hold the optic in place when not under vacuum;
• The six #8-32 holding screws should be changed to four #10-24 s.  Orient these equidistant and with lower edge parallel to the table. No helicoil.;
• Assembly drawing says o-ring part number #2-130 but description is is #2-223 (which is the next size class up). Choose one. The both are good in terms of groove size but maybe choose the smaller one;
• Include screw part in assembly table;
• It would be best practice to include a Teflon gasket on the flange side of the assembly between window and metal. Probably a thickness of 0.090".  Even if there is still an air gap left in the design dimensions. If the tolerances are a bit off then having a soft plastic surface is an good idea, it gives the optic something to rest against that won't apply hard localized stress points;
• Place centering points on assembly drawing of the front face for two windows. Also, add center line on Detail B view; and
• The two retaining rings are too close together.  Move spacing from 1.5" to 1.662".  This will mean the beams are not centered on the windows but will give some more space between the retaining rings.

Many of the finishing and tolerance parameters don't matter so much.  Over toleranced parts will cost more.  The only place where this might need to be careful is in the spec of the oring grove where vacuum is actually being held back, the 16 Ra in the drawing is fine.

Marco rubber and plastics have a good summary page for best practice design parameters, see: O-Ring Groove Design Directory.

For reference I have attached original schematic to this post.  Koji sent me the link to the original document that can be found here: https://labcit.ligo.caltech.edu/~rana/dale/Length_Sensing_and_Control/LSC_Photodiode/Version_B/D980454-01.pdf
 

This version of the document includes page 1 with the pinouts for the power connector.

To develop an accurate physical model to be used as a testing ground for the machine learning controls to be implemented in the system, all parameters (material and dimensional) must be known as accurately as possible. The parameters A, k as in https://nodus.ligo.caltech.edu:8081/CTN/2191 are not known very accurately so a measurement of the time constant (time taken for the can to naturally cool to 1/e of its initial temperature) was attempted in order to gauge how well the model matches with experiment.

Several previous measurements of the time constant were undertaken, but results varying from 2-5 hours were obtained https://nodus.ligo.caltech.edu:8081/CTN/1728 and therefore, to further investigate what was going on another such experiment was undertaken.

Experiment:

On 23rd May, Andrew and I surrounded the assembly of the vacuum can with aluminium foil (as seen in the picture). A setpoint of 45 C was chosen on 23rd May at 20:24, then on 24th May at 15:40 after the set point was reached the heater and PID were turned off. The evolution of the system was recorded.

The steady state temperature obtained was 45.792 0.005 C, where the uncertainty is calculated from a fit of steady state data as seen in the attached figure. A small region before cooldown was used for an estimate of the rms value for the noise obtained by subtracting a polynomial fit (10th order) to detrend the data. 

The region of the data corresponding to the cooldown was fitted with an exponential decay using scipy.optimize.curve_fit() with:

 
The following are the parameters from the fit:

The uncertainty here is taken from the square root of the correspoding covariance matrix for the fit parameters. This seems very unreliable given the unreasonably large uncertainties in a,c and relatively tiny uncertainties in b,d, even though visually the fit seems good.

According to the fit, the time constant should be 5.610 0.005 hours. But there seems to be many issues with the model including the large uncertainty and the very large  value that was calculated from the fit.

It seems like this model is an inaccurate description of the system at this level of sensitivity of measurement. The exponential decay curve does not even visually appear to fit the data to the level of the calculated rms noise value. This can be seen even in previous such experiments (as seen in https://nodus.ligo.caltech.edu:8081/CTN/1728). The possible reasons for this may be that the simple conduction model of the vacuum can may be leaving out conduction or radiation through other significant channels, the gradient across the foam is not linear at all time steps (as is assumed in the conduction equation), the geometrical effects of the foam and can may be more significant than is assumed, or the inner components of the can may be responsible for significant heat transfer. The answer to this may be evident from performing more complex models to fit the data.

The jupyter notebook can be found at https://github.com/CaltechExperimentalGravity/NonlinearControl/tree/master/TemperatureControl/Data/20180518_CoolDownTestVacCan

Hi shruti,

You might have more luck with scipy.curve_fit if you try taking the log of the y data, so log(y) = log(a * exp(-(t-c)/b) + d ) = some simpler expression, and create a new function around your simpler expression.

We expect the vacuum can to have multiple time constants since there are multiple sources of heat loss. Don't worry too much about fitting, just tune the model so that it matches the observation. i.e. the model should have more than 1 conductive cooling path. Once it is accurate to ~20-30%, better to focus on the feedback than the tuning of the model.

Redesigned D1400384-v2 is ready and uploaded on dcc. PCB Layout, circuit schematic, Gerber files, LISO analysis, front panel and full BOM are attached.
The circuit is kept the same except for replacing LT1128 with OPA827 in the transimpedance amplifier stages and adding few OMIT and ZERO resistances to have more options for using the circuit. Following are the notes (also present as ReadMe.txt in the .zip file in dcc):

Notes:
1) To not have whitening option, do not populate IC5, U8, U9 and their peripheral resistors and capacitors and put a 0 Ohm resistor (Jumper) at R20 and R26

2) To have fixed transimpedance, do not populate IC8 and its peripheral resistors and capacitors and populate R7 and R22 with the desired transimpedance.

3) In case of switchable transimpedance, R11=R15=100 and R14=R28=10k are the two options in present design. These can also be changed according to choice. Mark the front panel accordingly with the choices.

4) For low transimpedances, it is better to replace U5 and U7 with LT1128 instead.

The board design is finished and all the files have been uploaded on dcc (LIGO-D1800214-v1). PCB Layout, circuit schematic, Gerber files, LISO analysis, front panel and full BOM are attached. There are two copies of ISS circuits on the single board and they will be mounted on rack with a 2U front panel.

Notes:
1) The Noise performance analysis is present in the LISO_and_LTSpice_Files folder.
2) The present values in stage 4 of the circuit has cavity pole neutralization for a pole frequency of ~40kHz. To make this different, change the capacitor C22 and C53 accordingly.
3) The resistors R9 and R14 in Stage 0 can be populated to give an overall gain to the circuit.

I and Andrew measured transfer functions of each stage of PMC Servo box using Agilent 4395A and HP 41800A probe. Attached are the measured results overplotted with LISO analysis. It seems the PMC Servo box is functioning as it is designed for. Note that in the plots, at higher frequencies there are flat or increasing magnitude regions. These are just because of measurement reaching noise floor. We check independently with sweeps at increased power for higher frequencies and the curves match LISO analysis more or less.
Next, we connected the PZT to the servo box and measured the transfer function with PZT. Again this came out as expected with a good one pole low pass filter behavior as the circuit is designed for. The transfer function (5th plot in the file) keeps going small with a linear slope well up to 10 MHz (measured till this point). So everything checks out fine with he PMC servo box.
But as we lock the servo and measure the closed loop transfer function of the system, we see a strange flat region above 400Hz. Even after the low pass filter with PZT, in a closed loop, it looks like something is actively canceling the pole above 400Hz. But there is nothing else active in the loop. The 6th plot is the measured open loop transfer function (calculated from closed loop transfer function). Note that this flatness is not noise floor or saturation of any kind. We verified this by changing source power and the frequency response at output changes proportionately to this source power change (eg if source power is changed from 5mV to 50mV, the frequency response at the output at 500Hz changes from 4.5mVpkpk to 45mVpkpk), maintaining the flat behavior above 400Hz.
The open loop transfer(G)  function was calculated from closed loop transfer function (C) using this formula:
                                                G = 1 + (1.216)/C
where 1.216 is the measured flat response of the buffer in between the test input (FP2Test) and test output (FP3Test).
So we would like if someone can give some suggestions with this.
Addition: Attaching up-to-date schematic

post up-to-date schematic

Update Thu Aug 30 15:13:48 2018: now added to elog 2228

[awade and anchal]

ANCHAL ADD UPDATED TF PLOTS AND PART NUMBERS

For the PMC electronics: this flattening of the TF above 400 Hz was because the on board LP filtering in the RF demodulation stage (LP filter after the mixer) was letting too much RF through in the pass band.  This is a design floor of this particular board that a 1/f^2 filter (made with discrete components) was used instead of a all-in-one 1/f^4 chip.  With too much RF going into the LT1028 there is some kind of saturation problem which causes a leveling off above 400 Hz when taking the TF.

To fix this we put an external minicircuits mixer (model?) followed by a 1/f^4 minicircuits (model?) filter with a corner of 1.9 MHz -- with 50 Ω termination -- into the FP1test point (J3).  At the modulation frequency of 21.5 MHz this should give an attenuation of order -42 dB, much better than before.

[INSERT OLTF]

With the demodulation performed with eternal components the open loop transfer function behaved as expected with the 10 Hz pole setup by the output resistor and PZT working all the way out to very high frequency.  As a result the loop is now stable and working as expected.

We installed new RFPDs with components s.t. they are tuned near 35.5 MHz. We tuned the inductor L1 and capacitor C40 to bring the peaks and notches to desired values. Attached are TFs and dark noise measurements of these two photodiodes. The transfer function measurements are adjusted with the transfer function of the cables up to the photodiodes. The photodiodes were blocked with a beam dump during measurements and are supplied with +-18 V through the new power strip under the table. In the analysis, 2000 Ohms of DC transimpedance is assumed for TFs. Note: This was a wrong assumption. So do not trust the y-axis. This is a transfer function measurement through test-in ports which could be wrong.
Following are some measured numbers:

South Path:
SN 010
RFPD tuned to 37 MHz.
Notes:
Inductor L1 changed to 146-06J08SL
+5V regulator LM309H replaced with new one.
Measured Peak at 36.882 MHz
Measured Notch at 74.237 MHz

North Path:
SN 009
RFPD tuned to 36 MHz.
Measured Peak at 36.247 MHz
Measured Notch at 72.356 MHz

Conclusions:
Except for South RFPD being low Q (as it was written on it). the two RFPDs have similar response and same amount of dark noise. This should reduce the contributions from south path in the excess noise measured in the last BN measurement.

Please explain in much more detail how the RF transimpedance calibration was done.