I know the squeezing people often use the zero-span feature of a spectrum analyzer to produce their McDonalds plots, but why not just use a LC bandpass (e.g. a 1 MHz wide mini-circuit filter)?
The transimpedance amp could drive a mini-circuits amp which then drives the bandpass into a RMS->DC circuit (some diodes). Then you can plot it on a scope easily. Maybe not worth it for this first measurement.
I think the dark noise should be very close to the shot noise and also have the same transfer function.
Also, the room temperature fluctuations are probably large at low frequencies. What PSD for room temperature did you use?
OK...but how will the amplitude stabilization be done? How about a diagram showing the feedback loop and electronics?
We need to get the temperature data from the sensors that Aidan setup. I have got it by downloading a sketchy HOBO app onto my phone. In order for this to work you have to go stand next to the sensors and download via bluetooth. There is no "pairing" required - the app seems to recognize the sensors nearly immediately.
and please, Aidan and everyone else, please do not use double-sticky tape or glue or bubble gum to attach sensors in a clean optical setup. We're going to have to make a mess to remove these.
I've just downloaded more data (attached). The lab seemed sort of cold and the south HEPA output was blowing colder air than the north HEPA output.
I've changed the temperature from 70F to 73F at 7:45 PM today. Lets see how it goes.
Looks good! A few comments:
the temperature has stabilized, although the fact that there is a sign error in one of the HVAC actuators means that there is a big delta T in the lab. I will contact facilities again.
the change on April 24 is me adjusting the setpoint on the south wall from 70 to 73F.
the python script in the attachment will now scrub the data file for bad lines and make this kind of plot. IT works on .csv files, but none of the other HOBO formats.
Modesto came back this morning. We were still getting 1 hot blower and 1 cold blower.
As it turned out, although hot water was reaching the south heating coils input, it was not getting into the actual heating duct (as we could tell from touching those coils).
Modesto tapped the actuation valve with a hammer and it seemed to loosen it up. We will record temperature over the weekend.
Also we noted that the damper valve near the south-east vent was adjustable, but had no effect on the output air velocity so we assume its busted inside - Modesto will crack it open to investigate.
We also reduced the flow of hot water to the north blowers since they always seemed to be hotter. This may balance the actuators somewhat.
Modesto also replaced the thermostat on the west wall, but no real effect. I have adjusted the setpoint from 74 to 71F around 1800 today.
fixed (?) HVAC problems (~1700 Monday) by better attaching a pneumatic hose that actuates the hot water flow for the south blowers
update on Tuesday: problem not fixed. Bumping the actuator there makes the thing work for a few minutes, but then it freezed again.
I've removed the Moku and its iPad from the lab to see what its up to noise-wise.
Oh no! We've lost all of the low frequency data (which is the whole point of this experiment) and all of the contributing noise sources to the circuit!
always keep the start frequency at 0 Hz no matter what the span, as I was showing you yesterday in the lab. Otherwise, the bin centers end up in weird places.
what about attaching a crane to the ceiling on one of the supporting beams?
I'm pretty sure that the OP27 data is still not right. You should use the small binwidth and larger # of averages as we talked about earlier this week. In the elog, you should give the PSD parameters.
This is a summary of some information on types of solder and their usefulness.
Summary: use the 63/37 Sn/Pb solder from Kester. It is eutectic and has a low melting point so that your opamps won't get damaged.
We want our solders to be "eutectic" so that it goes from the liquid phase directly to solid with no intermediate slurry. This makes a reliable (and nice looking!) solder joint.
The tin-lead solder is a good combo.
we'll suffer if we use a oil based pump long term - please find and order a dry pump to back this turbo. Not only is it bad for the vac chamber, its bad for other optics in the lab,
don't we also want to record the dark noise spectrum as a function of T and V_Bias ? I would guess that the dark noise doesn't always scale with dark current at low frequencies since its probably more like a random walk than shot noise.
would be easier to achieve there with higher laser powers and higher cavity finesse.
But I haven't attempted that here as we do not know our NPRO PZT's resonance features yet.
I don't know why it would be easier to have higher finesse with longer arms. Something about beam size???
The NPRO PZT TF's are all in the 40m elog - there are many measurements of TF made over the past 10 years. Its like Raiders of the Lost Ark - you have to believe its there while searching.
I was thinking about getting this new current pre-amp from NF:
It seems to have a good noise performance and has a built in low pass filter and also a remote interface.
The FEMTO seems less fancy, but their noise performance is actually 2-3x better.
FEMTO DLPCA200 low noise preamp (brand new)
Keithley Source Meter 2450 (brand new) => Returned 11/23/2020
were brought to the OMC lab for temporary use.
doesn't seem so, but they sell this one:
which has a USB interface and pretty good voltage noise spectrum
is the x-axis in units of seconds? I think if we are clever, we should be able to look at a couple of the thermal time constants and figure out where the heat leaks are.
To aid in taking photos of these diodes, I put a USB microscope on Anchal's desk - you can grab it from there. I use it with mac Photo Booth, but it should be easy to use with any camera application.
Also, I recommend buying a macro lens(es) for cell phones from Amazon or B&H. Label them with the QIL lab sticker so they don't disappear.
Looks very clear, thanks. I guess the next thing to do is
Note that the back panel connectors are Triax, not the usual Coax.
exp plot tip: If you use the "grid" feature of matplotlib and plt.semilogy(), the exponentials will look like straight lines, so we can just read off the time constants with a ruler.
Also, as we talked about earlier today, we should make some analytical estimates for the various heat loads, and also put them into the model.
For protecting from radiation, all of the surfaces which are NOT shiny-polished should get wrapped in something shiny (UHV Al foil, with the shiny side out).
I suggest wrapping with foil:
I attach here a photo of the radiative shielding of a Purple Pepper Plant (PPP), to reduce the radiative coupling to the environment. This prevents the soil from drying out in the sun so fast.
I'll be curious to see the results of Radhika's thermal model - I am suspicious of this thermal strap contact to the base plate. It would be good if we could instead make a copper mating plate:
I think the asymptotic temperature in the model is missing the data. i.e. the steady state temperature should match up, but the recorded data terminates too soon. Probably should figure out what the missing term is.
Can you post the covariance matrix of your fit so that we can see what the fractional errors are on the physical parameters? (i.e. construct the fit function so that physical parameters which are unknown are the fit parameters.)
We should not have a bench power supply installed permanently. Can you install a Sorensen in that rack or use one of the nearby ones?
you can put these in the GIT repo for the QIL Cryo tests that Radhika set up. Otherwise, they'll get lost. And we should probably change autorun to a .py script and document these in the README on the repo.
The attached files are the scripts used to take data during the PD temperature cycling/testing and to retrieve and analyze data after the fact.
its worth looking into how fiber optic mode cleaning actually works:
In order to get a lot of cleaning you have to have a clean beam to begin with. There's a way to pre-clean by putting the laser output into a pinhole before coupling int othe single-mode fiber. Also, use a ~40-50m fiber to make sure the mod-mistanatched beam actually goes into the cladding rather than recombine into the Gaussian beam.
agreed. You should put into Voyager chat and cryo/ET slack questins about 2 um beam profiling. We'll want it for anything 1.8-2.1 um.
Definitely. I think the lack of beam profiling/imaging equipment is something we want to address too. We will waste a lot of time in Mariner if we can't profile our beams.
IT would also be good if you could plot the RMS noise around 10 Hz and 100 Hz as a function of the bias and temperature, so we can see what the trends are. And how about post the data and scripts to the elog so we can munge data later?
thats looking good
You should try to use either the corner or getdist packages to plot the 'corner' plots commonly used when showing correlated posteriors (cf. https://emcee.readthedocs.io/en/v2.2.1/user/line/) so that we can see what's up with the other uncertainties
This week I have been working towards a Markov-chain Monte Carlo (MCMC) approach for fitting Megastat cooldown data and obtaining estimates on various emissivity values. I started with a simple model, only simulating the radiative cooling from the inner shield to the test mass. I supply the test mass and inner shield temperature data, and the emissivity of Aquadag (coating both the TM and inner surface of IS) as the only fit parameter. I am using Stan for the modeling, and Attachment 1 is a copy of my stan model.
The results of the simple model are in Attachment 2. The emissivity of Aquadag is estimated as a gaussian centered around 0.7. I am still determining whether this result is "real", or if the simulation is simply returning my prior. I will look into this more before adding complexity to the model.
Is there a units issue? 50 kg is a mass, but not a force.
For one surface resting on another, the force of the contact is the grativational force of the top surface. There's an implicit factor of that cancels out from the ratio, so it becomes a ratio of masses. The heat conductance listed for a 50 kg Al-Al pressure joint is interpreted as the heat conductance for a force of .
I think its least confusing to just replace 50 kg g with 500 N. Writing 50 g can be misleading, it seems like 50 grams.
In the plots comparing data and models, can you use the legend to indicate which is which? e.g. use dots for data and solid lines for models, and then label them as that in the legend. Also nice to include error bars on the temperautre measurements; I think there's a python way to plot this as a shaded region as well.
I've been assuming that the inner shield can be treated as a point mass, but perhaps the thinness make a significant delay between the temperature of the cold plate and the inner shield during the initial cooldown.
Could you model the cold shield to estimate what the temperature gradient would look like during the rapid cooldown? Not full 3D, but something approximate that takes into account the conductivity, thinness, and heat capacity.
we don't really need Pneumatic legs. How much for rigid legs?
I think this is a nice debugging find. Its not very robust to use the workstations as 24/7 script machines (as we have found out over the years).
Best is to install a conda env on the main framebuilder machine, and run the perpetual scripts there in a tmux session.
Once its all sehup, update the ATF Wiki with a description of haw its done. Workstations crash when users do stuff, so its better if the data gettin script can run as a system service (e.g. systemctl, etc)
that's good. Can you from these models estimate what the uncertainty will be on the emissivity? i.e. by MCMC or otherwise rather than eyeballing
I've modeled the cooldown of a 2" diameter and 4" diameter Si wafer in Attachments 1 and 2, using the current Megastat model and previous cold head temperature data. The model includes heat leaking into the inner shield enclosure from an aperture, which we currently observe in Megastat cooldowns. (Note how the wafer cools down much faster than the current test mass, due to the very tiny volume.)
I'd recommend drilling a through hole and using a nut, rather than tapping as I suggested earlier. The shield is not thick enough to have many 4-40 threads.
Is it better to have the inner shield inner surface low or high emissivity? \Depending on the effect on the MCMC uncertainties, we may want to make everything black, including the cold plate. i.e. we could mount something like black aquadag tiles on the cold plate.
that's a good start, but we want the cooldown to be fast, so what we would need from the model is for you to change some parameters and find out what kind of possible physical changes we can make to make the cooldown fast. i.e. change some parameters and see what configurations would give a fast cooldown, and then we can discuss which of those is the easiest.
I modeled the cooldown of a 2" Si wafer in Megastat with cooling from a) the existing cryocooler and b) LN2. The only difference between the two models is the temperature that the cold end of the copper bar "sees" - in the cryo-cooler case, I've used cold head temperature data from a previous cooldown; in the LN2 case, I've used 77K. In Attachment 1, the cryo-cooler models are solid lines and the LN2 models are dashed.
It is clear from the plot that the cold head gets ~30K colder than LN2 at 77K. This explains the discrepancy between the inner shield models for the two cases at steady state. While the initial temperature rates of change are greater in the LN2 case, the cold head crosses the 77K line in roughly 5 hours.
Does the true cold-head temperature follow the model? Or is it less good than the model?
From the plot, it seems like the slop is ~50 Ohms / 120 K. So a difference of 5 K corresponds to ~2 Ohms.
It could be that your measurement of the resistance is off slightly due to the lead resistances, etc. Are you using a 3-wire or 4-wire probe?
You can get a higher accuracy relative calibration by dunking all the RTDs together in ice water and then in boiling water.
Today we moved some of the optics away from the blue box to make room for the mode matching.
We also found a 1" diameter optic in one of the cabinets with a transmission of 0.81 % (S-pol). This makes a nice match with the 0.57% transmission of the 2" diameter output coupler and so its now installed. We'll later replace it with a good 2" diameter mirror.
We also moved all of the optics after the laser to align with the screw holes and are trying to make the beams go along the screw hole lines.
Alastair has remeasured the laser beam profile after the EOM and the PBSs using the WinCamD beam scanner. Results are being produced.
The script which cleans up the frame files (so that the disk doesn't become overful) was set to only delete files when the /frames/full directory was getting up to 99.7% of the full capacity. This is ridiculously close to the edge. We set it instead to be 95%. Here's the diff in the /target/fb/wiper.pl script:
fb0:fb>diff wiper.pl wiper.pl~
< $full_frames_percent_keep = 95;
> $full_frames_percent_keep = 99.7;
< $minute_frames_percent_keep = 0.2;
> $minute_frames_percent_keep = 0.005;
The DAQD process was spitting out core files and had also filled up the / partition on FB0. After deleting this the regular system processes were able to run. To check the disk space you can use 'df -h':
Filesystem Size Used Avail Use% Mounted on
224G 24G 189G 12% /
/dev/sda1 99M 28M 66M 30% /boot
tmpfs 1006M 0 1006M 0% /dev/shm
/dev/sdd1 1.4T 142G 1.3T 11% /frames/trend
/dev/sdc1 917G 707G 165G 82% /frames/full
fb1:/cvs 917G 104G 814G 12% /cvs
We found that although NTPD was running on FB0, it had been configured in some really screwy way. We used /sbin/ifconfig to remove the configuration for the other network devices (eth1, eth2, eth3) and set it so that FB0 only talks to the ATF martian network and the router. The router is now configured to NOT filter out the requests from FB0. Now the NTPD works and seems to be correctly fixing the computer's system time. There's still the issue that the ATF FE will change this time as long as the FE is running, but I guess the system clock will once in awhile get fixed when we restart the FE and NTP takes over.
Along the way, I also restored the Xinerama dual-head display on ws1. Alastair somehow believed that it had never been dual-head before, but in fact I elogged the procedure in September. Please don't do any auto-updates on any of these machines unless you know what you're doing. and are willing to fix things after breaking them.
I attached an image showing that I can, indeed, get real time data from one of the _DAQ channels of the gyro.
That's a weird plot. I think we want to see the HOM of the SBs, not the carrier. Or rather, we want to see both but maybe not on the same plot. Are these the SB HOMs?
If so, I think its fine to have an 8th order HOM of a SB to have the same phase shift as the TEM00 SB. I say that the frequency should be between 21 and 45 MHz.
I had a rather lengthy discussion with Aidan, Frank and Koji about this the other day. The resonance peaks are resonances of the cavity for some higher order mode. They can be excited by either the carrier or the SBs impartially. This plot is equivalent to the ones that are in other elog entries (like John's from last year), but it is much less cluttered. Those plots show you exactly how far each higher order mode is from resonance for the carrier and each sideband, whereas this one shows you the resonance frequency of each cavity HOM, along with the frequency of the carrier and SBs, from which you can directly infer how far each one is from whichever mode peak. I think this is a clearer way of doing it, as it emphasizes that the resonances belong to the cavity, not to the light entering it.
No, I disagree. There should be a plot showing how the HOMs of the SBs show up in the cavity. A TEM01 of the +SB will be in a different place than the -SB TEM01 or the CR TEM01.