Since I am finishing my job at the lab, I have stored all my electronics in a box (attachment 1) and placed it under the table in the control room where some other electronics are stored. The box contains the heater circuit box, two temperature sensor boards, one temperature sensor, a short power cable and +/- 15V supply cables. In the lab I left the wires for the current setup and tied them down to the wall so that they aren't in the way (attachment 2). I left the can as is and the other temperature sensor is still attached to the inside of the can. I have labeled the wires going from the sensor as 'in' and 'out'. I've also left the wires for the heater there as well (attachment 3). I turned off the PID control and deactivated the tmux session on megatron.
Thanks to Rana and the LIGO team for giving me the opportunity to work at the 40m on this project with the seismometer.
I'm running a comsol job on optimus in a tmux session named cryocavs. Should be done in less than 24 hours, judging by past durations.
I have connected a 4TB disk to chiara via a USB-SATA adapter. This disk has been recognized as /dev/sde. A GUID Partition Table (GPT), not MBR was made with gdisk to make a partition with the size beyond 2TB.
I tried to use "dd" to copy /home/cds (/dev/sdb1) to /dev/sde1, but failed. The copy was done (taking ~12h) and the partition was not recognized as a complete filesystem.
So I decided to use rsync instead.
sudo mkfs -t ext4 /dev/sde1
sudo mkdir /media/usb4g
sudo mount -t ext4 -o rw /dev/sde1 /media/usb4g
sudo rsync -a --progress /home/cds/ /media/usb4g
14:33 Copied 33G/1831G
14:38 Copied 36G/1831G
17:02 Copied 365G/1831G (~2.2GB/min)
01:18 Copied 1449G/1831G (~2.2GB/min)
> sent 1907955222607 bytes received 126124609 bytes 37010956.31 bytes/sec
> total size is 1907271994803 speedup is 1.00
Here's a Finesse modeling of what we're expecting to observe with this test. It uses Gautam's base model of the 40m IFO with appropriate modifications for the needed configuration.
The idea is to lock the IFO in the SRMI configuration, with the phase-locked AUX beam injected from the AS port. The AUX beam is imprinted with AM sidebands and slightly misaligned relative to the SRC so as to transfer power into HOM1. The RF network analyzer provides the drive signal for the AOM, and its frequency is swept to coherently measure the transfer function [reflected AUX beam / drive]. The reflected AUX beam is sensed by the AS110 PDA10CF.
It is also possible to drive PM sidebands as Koji suggests, but the squeezer group has encouraged using AM for practical advantages. The SNR with AM is a bit higher (less power lost into harmonics at large modulation index), there is a broadband AOM already available aligned to the SQZ beam at LLO, and there is also concern that driving strong PM could interfere with the SQZ control loops.
Attachment #1 shows the expected response to swept-AM in SRMI. Resolving just the FSR and the first-order mode splitting is sufficient to extract the SRC Gouy phase.
Since the 40m has not been opearted in SRMI since ~2016 (last done by Eric Q.), Gautam believes it may take some time to relock this configuration. However, the modeling indicates that we can likely obtain sufficient sensitivity in DRMI, which would allow us to proceed faster. Attachment #2 shows the expected response to swept-AM in DRMI. The PRC leakage signal turns out to be significantly smaller than the SRC reflection (a factor of ~30 in amplitude), so that the signal still retains its characteristic shape to a very good approximation. The tradeoff is a 10x reduction in SNR due to increased PSL shot noise reaching AS110.
Based on this, we should proceed with DRMI scans instead of PRMI next week.
The PRC FSR is, of course, very close to twice of our f1 moudlation frequency (11MHz x 2 = 22MHz) .
I still don't understand what response the measurement is looking for. I understood the idea of using the subcarrier as a stablized carrier to the PRC with a certain freq offset from the main carrier. I suppose what was swept was the AOM modulation frequency (i.e. modulation frequency of the AM applied to the subcarrier). If that is the case, the subcarrier seemed fixed at an arbitorary frequency (i.e. 50MHz) away from the carrier. If one of the AM sidebands hits the PRC resonance (i.e. 22, 44, 66MHz away from the main carrier), you still have the other sideband reflected back to the AS. Then the RF signal at the AS is still dominated by this reflected sideband. I feel that the phase modulation is rather suitable for this purpose.
If you are talking about ~MHz AM modulation by the AOM and scanning the PLL frequency from 1MHz to 60MHz, the story is different. And this should involve demodulation of the AS signal at the AM modulation frequency. But I still don't understand why we don't use phase modulation, which gives us the PDH type signal at the reflection (i.e. AS) port...
[Jon, Gautam, Johannes]
We did the following today:
This measurement seems like a fine candidate to trial the idea of looking for the FSRs (and in general, cavity resonances) of the PRC in the phase of the measured TFs, rather than the amplitude.
I did the measurement with the BeatMouth open today. Main changes:
So neglecting asymmetry in the branching ratio of the fiber beamsplitter, the asymmetry between the test PD optical path and the reference PD optical path is a single fiber mating sleeve in the former vs a collimator in the latter. In order to recover the expected number of 409 V/W for the Menlo PDs, we have to argue that the optical loss in the test PD path (fiber mating sleeve) are ~3x higher than in the NF1611 path (free space coupler). But at least the X and Y PDs show identical responses now. The error I made in the previously attached plot was that I was using the 20dB coupled output for the X PD measurement .
Revised conclusion: The measured optoelectronic response of the Menlo PDs at 10s of MHz, of ~130 V/W, is completely consistent with the numbers I reported in this elog. So rogue polarization is no longer the culprit for the discrepancy between expected and measured RF beatnote power, it was just that the expectation, based on Menlo PD specs, were not accurate.#2 of the linked elog seems to be the most likely, although "broken" should actually be "not matching spec".
While killing time b/w measurements, I looked on the ITMY optical table and found that the NF1611 I mentioned in this elog still exists. It is fiber coupled. Could be a better substitute as a Reference PD for this particular measurement.
I will repeat the measurement tomorrow by eliminating some un-necessary patch fiber cables, and also calibrating out the cable delays.
Aim : To develop a neural network on simulated data.
I developed a python code that generates a 64*64 image of a white Gaussian beam spot at the centre of black background. I gave a sine wave of frequency 0.2Hz that moves the spot vertically (i.e. in pitch). Then I simulated this video at 10 frames/sec for 10 seconds. Then I saved this data into an hdf5 file, reshaped it to a 1D array and gave as input to a neural network. Out of the 100 image frames, 75 were taken as training dataset and 25 as test data. I varied several hyperparameters like learning rate of the optimizer, number of layers, nodes, activation function etc. Finally, I was successful in reducing the mean squared error with the following network model:
I have attached the plot of the output of neural network (NN) as well as sine signal applied to simulate the video and their residula error in Attachment 1. The plot of variation in mean squared error (in log scale) as number of epochs increases is given in Attachment 2.
I think this network worked easily since there is no noise in the input. Gautam suggested to try the working of this network on simulated data with a noisy background.
Oplev sums of 240 days.
Since there have been various software/hardware activity going on (stack weighing, AUX laser PLL, computing timing errors etc etc), I decided to do a check on the state of the IFO.
The cabeling was cleaned up a little bit yesterday morning. The upper back side is still massy.
Using the numbers from the sensing measurement, I calibrated the measured in-loop MICH spectrum from Tuesday night into free-running displacement noise. For convenience, I used the noise-budgeting utilities to make this plot, but I omitted all the technical noise curves as the coupling has probably changed and I did not measure these. The overall noise seems ~x3 higher everywhere from the best I had last year, but this is hardly surprising as I haven't optimized anything for low noise recently. To summarize:
I will do a more thorough careful characterization and add in the technical noises in the coming days. The dominant uncertainty in the sensing matrix measurement, and hence this free-running noise spectrum, is that I haven't calibrated the actuators in a while.
I finally analyzed the sensing measurement I ran on Tuesday evening. Sensing responses for the DRMI DOFs seems consistent with what I measured in October 2017, although the relative phasing of the DoFs in the sensing PDs has changed significantly. For what it's worth, my Finesse simulation is here.
Jon spent some time trying to align the AUX beam to the SRC today, I got to the game kind of late so maybe others can add more detail.
The AUX beam that is reflected by the SRM looks terribly misshapen - it is quite elongated in vertical direction. Unfortunately I didn't snap a picture of it - anybody? It seemed at first as if this could be clipping - but after confirming the alignment of the AUX beam with the PSL output beam with aligned SRM, a slow dither of the SRM just moved the ugly pattern on the AS camera with no change to its shape - so clipping is unlikely. I'm now thinking that this is just the output beam of the fiber coupler after propagating ~15 meters to the SRM and back - even though this aspheric lens triplet coupler is supposed to be super-duper. I found that if I loosen the fiber slightly and pull it back just a bit at least the spot on the AS camera becomes nice and round - so maybe the fiber just doesn't sit well in this collimator? Not sure why that would be. I checked the fiber tip with the microscope, and while there was some gunk present, the central region and the core were clear (still cleaned using the fiber cleaning kit, which got rid of the debris). Either way, before switching to a different collimator I think we should give the Guoy phase measurement a shot - after all there was plenty of RF signal present on both AS110 and the PDA10CF placed at the YEND.
Looking for rogue beams on the AS table, I started placing some beam dumps. There was one particularly strong source of stray beams - a lens that was labeled with KPX094AR.33_F100. It became apparent after alignment efforts to the IFO had moved the AUX beam signifcantly off-center on this lens. According to the label it should have an AR coating for 1064nm, however judging by the amount of reflected light, it was certainly NOT AR-coated for 1064nm. I replaced it with a bi-convex f=100mm lens with confirmed AR-behavior.
The AUX laser is currently shuttered.
Per our Wednesday meeting, some items to work on are
All optics have been re-aligned. Jon/Johannes will elog about the work today.
Bad wording, sorry. Should have been channels in excess of ETMX controls. I'll add the others to the list as well.
Updated channel list and wiring diagram attached. Labels are 'F' for 'Front' and 'R' for - you guessed it - 'Rear', the number identifies the slot panel the breakout is attached to.
We have 6 of these boards now in cabinet E7
I wired all 32 channels going to the AA board directly to the ADC as described in the previous log. However, instead of using the old AA board and bypassing the whole circuit, I just used a breakout board as is shown in the first attachment. I put the board back in the rack and reconnected all of the cables.
The seismic BLRMs appear to be working again. A PSD of the BS seismometers is shown in attachment 2. Tomorrow I'll look at how much the ADC alone is suppressing the common mode 60 Hz noise on each of the channels.
Steve: 5 of ADC DAC In Line Test Boards [ D060124 ] ordered. They should be here within 10 days.
I think this is because /cvs/cds is getting too big. lsblk reveals:
NAME MAJ:MIN RM SIZE RO TYPE MOUNTPOINT
sda 8:0 0 465.8G 0 disk
├─sda1 8:1 0 446.9G 0 part /
├─sda2 8:2 0 1K 0 part
└─sda5 8:5 0 18.9G 0 part [SWAP]
sdb 8:16 0 2.7T 0 disk
└─sdb1 8:17 0 2T 0 part /home/cds
sr0 11:0 1 1024M 0 rom
sdc 8:32 0 1.8T 0 disk
└─sdc1 8:33 0 1.8T 0 part /media/40mBackup
sdd 8:48 0 1.8T 0 disk
└─sdd1 8:49 0 1.8T 0 part
I believe one of sdc or sdd is connected via SATA while the other is an external USB drive. Maybe we have to get bigger backup disks, but this may be a huge pain to setup as it will involve taking chiara down. Actually, now that I check the backup log, seems like backup is executing successfully - not sure if this is due to my unelogged mounting of sdc (using sudo mount /dev/sdc1 /media/40mBackup) last week, or if this is some LDAS backup. But in any case, seems undesirable that sdb1 is larger than sdc1 or sdd1.
2018-06-06 07:00:01,086 INFO Updating backup image of /cvs/cds
2018-06-06 07:00:01,086 ERROR External drive not mounted!!!
2018-06-07 07:00:01,147 INFO Updating backup image of /cvs/cds
2018-06-07 07:00:01,147 ERROR External drive not mounted!!!
2018-06-08 07:00:01,244 INFO Updating backup image of /cvs/cds
2018-06-08 08:23:32,939 INFO Backup rsync job ran successfully, transferred 316870 files.
2018-06-09 07:00:01,465 INFO Updating backup image of /cvs/cds
2018-06-09 07:12:11,865 INFO Backup rsync job ran successfully, transferred 1926 files.
2018-06-10 07:00:01,842 INFO Updating backup image of /cvs/cds
2018-06-10 07:12:28,931 INFO Backup rsync job ran successfully, transferred 1656 files.
2018-06-11 07:00:01,294 INFO Updating backup image of /cvs/cds
2018-06-11 07:06:14,748 INFO Backup rsync job ran successfully, transferred 1664 files.
2018-06-12 07:00:02,081 INFO Updating backup image of /cvs/cds
2018-06-12 07:07:36,775 INFO Backup rsync job ran successfully, transferred 1870 files.
2018-06-13 07:00:02,194 INFO Updating backup image of /cvs/cds
2018-06-13 07:08:37,356 INFO Backup rsync job ran successfully, transferred 1818 files.
2018-06-14 07:00:01,753 INFO Updating backup image of /cvs/cds
2018-06-14 07:01:43,270 INFO Backup rsync job ran successfully, transferred 1744 files.
Local backup on chiara seems not working since Nov 19, 2017.
2017-11-18 07:00:01,504 INFO Updating backup image of /cvs/cds
2017-11-18 07:03:00,113 INFO Backup rsync job ran successfully, transferred 1954 files.
2017-11-19 07:00:02,564 INFO Updating backup image of /cvs/cds
2017-11-19 07:00:02,592 ERROR External drive not mounted!!!
Jon is doing some characterization of the AUX laser setup for which he wanted only the prompt retroreflection from the SRM on the AS table, so the PSL shutter is closed, and both ITMs and ETMs are misaligned. The prompt reflection from the SRM was getting clipped on something in vacuum - the ingoing beam looked pretty clean, but the reflection was totally clipped, as I think Johannes aligned the input beam with the SRM misaligned. So the input steering of the AUX laser beam into the vacuum, and also the steering onto AS110, were touched... Also, there were all manner of stray, undumped beams from the fiber on the AS table Jon will post photos.
Before we began this work, we found that c1susaux was dead so we rebooted it.
Do we really have 2 free ADC channels at EX now? I was under the impression we had ZERO free, which is why we wanted to put a new ADC unit in. I think in the wiring diagram, the Vacuum gauge monitor channel, Seis Can Temp Sensor monitor, and Seis Can Heater channels are missing. It would also be good to have, in the wiring diagram, a mapping of which signals go to which I/O ports (Dsub, front panel BNC etc) on the 4U(?) box housing all the Acromags, this would be helpful in future debugging sessions.
its painful, but you and I should probably take these out, bypass the switches and use them with fixed gain; the 'Reed Relay' attenuators are not a good part for this app.
The historical problem is that they tend to self oscillate with full gain because they had 2 MAX4106 in series which couple to each other in the bad way --- need to remove one of them and set the gain of the other one to 10.
The unfortunate discovery today was that the attenuator switches on the IMC WFS heads are actually assigned to individual segments, and they are active. That means that we have been running the WFS with an uneven gain setting.
don't use IN_1/IN_2: recall pizza meeting from a few weeks back: use IN1/EXC + Al-Gebra
Measured loop TFs - PRCL is a big mystery. Used these to finalize loop gains.
I went through the wiring of the c1auxex crate today to disentangle the pin assignments. The full detail can be found in attachment #1, #2 has less detail but is more eye candy. The red flagged channels are now marked for removal at the next opportunity. This will free up DAQ channels as follows:
This should be enough for temperature sensing, NPRO diagnostics, and even eventual remote PDH control with new servo boxes.
Neither of the Menlo FPD310 fiber coupled PDs in the beat mouth have an optoelectronic response (V/W) as advertised. This possibly indicates a damaged RF amplification stage inside the PD.
I have never been able to make the numbers work out for the amount of DC light I put on these PDs, and how much RF beat power I get out. Today, I decided to measure the PD response directly.
In the end, I decided that slightly modifying the Jenner laser setup was the way to go, instead of futzing around with the PDFR laser. These PDs have a switchable gain setting - for this measurement, both were set to the lower gain such that the expected optoelectronic response is 409 V/W.
[Attachment #1] - Sketch of the experimental setup.
[Attachment #2] - Measured TF responses, the RF modulation was -20dBm for all curves. I varied the diode laser DC current a little to ensure I recovered identical transfer functions. Assumptions used in making these plots:
[Attachment #3] - Tarball of data + script used to make Attachment #2.
The unit mentioned in the x-axis was wrong. So I have remade the graphs. The point where frequency equals to zero is actually the frequency corresponding to the laser, which is in the range of 1014 Hz and it caliberated as zero.
The cavity scan data obtained from the Finesse simulation is attached here. Fig1 indicates the cavity scan data in the absence of induced misalignment. In that case only the fundemental mode is resonating. But when a misalignment is induced, higher order modes are also present as seen in Fig2. This is in the absence of surface figure error in the mirrors. Now I am trying to provide perturbations to the mirror surface in the form of zernike polynomials and get the scan data fom the simulation. These cavity scan data can be used to develop fitting models. Once we have a model, we can use it to analyse the data from the experimental cavity scan.
I want to use the Fiber Coupled laser from the PDFR system to characterize the response of the fiber coupled PDs we use in the BeatMouth. The documentation is pretty good: for a first test, I did the following in this order:
Seems like stuff is working as expected. I don't know what the correct setpoint for the TEC is, but once that is figured out, the 1x16 splitter should give me 250 uW from each output for 4mW input. This is well below any damage threshold of the Menlo PDs. Then the plan is to modulate the intensity of the diode laser using the Agilent, and measure the optoelectronic response of the PD in the usual way. I don't know if we have a Fiber coupled Reference Photodiode we can use in the way we use the NF1611 in the Jenne laser setup. If not, the main systematic measurement error will come from the power measurement using a Fiber Power Meter.
I have modified the code for frequency scanning and have made it completely command line enabled. The code is written in python. It is saved in the name "frequency_scanning_argparse.py". I have uploaded it to the Mode-Spectroscopy Github repository.
Inorder to use this code there are two ways.
1. We can mention the ' frequency' on which marconi need to work. Then it will change the marconi frequency to that perticular value.
eg: Type in the terminal as follows for changing the marconi frequency to 59 Mhz.
python frequency_scanning_argparse.py 59e6
2. Inorder to give a scan to the marconi frequency, provide the 'start frequency', 'end frequency' and the 'number of points' in between. This will be more conveniant when we want to run the scan in different ranges.
eg: Type in the terminal as follows for a start frequency of 59 Mhz, end frequency of 62MHz and number of points in between equal to 1000.
python frequency_scanning_argparse.py 59e6 62e6 1000
In both cases the code will show you the frequency of the marconi before we run this code and it will change the marconi frequency to the desired frequency.
Attachment #1 shows the measured PRCL loop shape. The blue line is meant to be the "expected" loop shape. While the measured loop shape tracks the expectation down to ~100 Hz, I cannot explain the shape below it. I am also not sure what to make of the fact that there is high coherence down to 10 Hz fron IN2 to IN1, but no coherence between EXC/IN2. I confirmed that the low-frequency boost filters were ON during the measurement. I don't understand how a pendulum TF + the digital filters we used can account for the shape below 100Hz.
gautam 11pm: After discussing with Koji, I conclude that the low frequency loop shape is consistent with the excitation amplitude being insufficient below 100 Hz. Coherence is good between In1/In2 because they are the same signal effectively - what we need is coherence between In1 and EXC, which isn't plotted. It is still strange that Coherence between In2/EXC is ZERO....
With Koji's help, I got repeatable and reliable DRMI locking going again tonight - this is with the AS path optics for the spectroscopy measurement in place, although the AUX laser remained shuttered tonight. Results + spectra tomorrow, but here's what I did:
As I have found before, it is significantly easier to get the locking going post 11pm - the wall Seis BLRMS don't look that much quieter at midnight compared to 10pm, but this might be a scaling issue. I'll do a quantitative assessment next time... Also, Foton takes between 25-45 secs to save an updated filter (timed twice today).
Today I made the led (1050nm) circuit inside a box as given in my previous elog. Steve drilled a 1mm hole in the box as an aperture for led light.
Resistance (R) used = 665 .
We connected a power supply and IR has been detected using the card.
Later we changed the input voltage and measured the optical power using a powermeter.
Since the optical power values are very less, we may need to drill a larger hole.
Now the hole is approximately 7mm from led, therefore aperture angle is approximately 2*tan-1(0.5/7) = 8deg. From radiometric curve given in the datasheet of LED1050E, most of the power is within 20 deg. So a hole of size 2* tan(10) *7 = 2.5mm may be required.
I have also attached a photo of the led beam spot on the IR detection card.
Koji's collection of Yend components put away. I cleaned up the Xend bench today.
Loadcells, leveling wedge mounts and related items placed under flowbench cabinet next to Guralp staff.
Steve mentioned two unlabelled optics were found at EX, relics from the Endtable upgrade.
These are now labelled and forked down on the SP table.
I worked a bit on recovering the DRMI locking again tonight. I decided to shutter the AUX laser on the PSL table at least until I figured out the correct locking settings. As has become customary now, there was a cable in the AS beampath (leading from the AS55 DC monitor to nothing, through the enclosure side panel, it is visible in Attachment #3 in this elog) which I only found after 30mins of futility - please try and remove all un-necessary cables and leave the AS beampath in a usable state after working on the AS table! In the end, I got several short (~3mins) stretches in tonight, but never long enough to do the loop characterization I wanted to get in tonight, probably wrong gains in one or more of the loops. In the last 30 minutes, the IMC has been frequently losing lock, so I am quitting for now. The AUX laser remains shuttered.
Per this elog, we don't need any AIOut channels or Oplev channels. However, the latest wiring diagram I can find for the EX Acromag situation suggests that these channels are hooked up (physically). If this is true, there are 12 ADC channels that are occupied which we can use for other purposes. Question for Johannes: Is this true? If so, Kira has plenty of channels available for her Temperature control stuff..
As an aside, we found that the EPICS channel names for the TRX/TRY QPD gain stages are somewhat strangely named. Looking closely at the schematic (which has now been added to the 40m DCC tree, we can add out custom mods later), they do (somewhat) add up, but I think we should definitely rename them in a more systematic manner, and use an MEDM screen to indicate stuff like x4 or x20 or "Active" etc. BTW, the EX and EY QPDs have different settings. But at least the settings are changed synchronously for all four quadrants, unlike the WFS heads...
Unrelated: I had to key the c1iscaux and c1auxey crates.
The unfortunate discovery today was that the attenuator switches on the IMC WFS heads are actually assigned to individual segments, and they are active. That means that we have been running the WFS with an uneven gain setting. The attached PDFs show that the signals with the attenuators on and off all at the same time, while the WFS servo output was frozen. A more annoying feature is that when some of the attenuators are on, this does not lower the gain completely. I mean that the attenuated channels show some reduction of the gain, but that is not the level of reduction we see when all attenuators are turned on. This RF could come from some internal RF coupling or some similar effect.
Moreover, the demodulation phases are quite off for most of the segments.
So far, the WFS is running with this uneven attenuation. We take time to characterize the gain and retune the demod phases and input matrices.
Oopss !! I made a mistake while taking the values from my notes. Sorry.
> The percentage error which I found out =[(analytical value - finesse value)/analytical value]*100
Yes, I this does not give us 0.70%
(3.893408 - 3.8863685)/3.893408 *100 = 0.18%
But any way, go for the fitting.
The percentage error which I found out =[(analytical value - finesse value)/analytical value]*100
But inorder to find the finesse value, I just used curser to get the central frequency of each peak and by substracting one from the other I found TMS and FSR.
The resolution was 6500 Hz. Thus, it seems that this method is not actually reliable. I am trying to find the central frequency of each mode with the help of lorentzian fits. I am attaching a fit which I did today. I have plotted its residual graph also.
I am uploading 4 python scripts to the github.
1. Analytical Solution
2. Finesse model- cavity scan
3. Finesse model- fitting
4. Finesse model- residual
Hmm? What is the definition of the percentage error? I don't obtain these numbers from the given values.
And how was the finesse value obtained from the simulation result? Then what is the frequency resolution used in Finesse simulation?
Why is this happening so frequently now? Last few lines of error log:
I fixed it by running the reboot script.
Hmm? What is the definition of the percentage error? I don't obtain these numbers from the given values.
And how was the finesse value obtained from the simulation result? Then what is the frequency resolution used in Finesse simulation?
Aim: To calibrate CCD of GigE using LED1050E.
The following table shows some of the specifications for LED1050E as given in Thorlabs datasheet.
The circuit diagram is given in Attachment 1.
Considering a power supply voltage Vcc = 15V, current I = 20mA & forward voltage of led VF = 1.25V, resistance in the circuit is calculated as,
R = (Vcc - VF)/I = 687.5
Attachment 2 gives a plot of resistance (R) vs input voltage (Vcc) when a current of 20mA flows through the circuit. I hope I can proceed with this setup soon.
As of now, I have made the codes needed to sweep the marconi frequency for taking the cavity scan data, the photo diode at the y-end is conected to the spectrum analyser already and I also have the finesse simulation of the Ideal Fabry-perot cavity. By seeing my last elog entry, Gautam suggested me that I need to take a different approach for estimating the FSR and TMS value from the Finesse graph. That is, by using least square fit models. Now I am trying to do that and get a better estimate of the error values. Based on my understanding I am dividing this project into various tasks.
1. Getting a better estimate of the error value by using least square fits. Also plotting a graph of frequency Vs mode number and finding the value of Free Spectral Range from its slop.
2. Inserting zernike polynomials to the Finesse simulation and with the help of least square fit, plotting the graph of frequency Vs mode number. Understanding the shifts from the Ideal graph we obtained from step 1. Using this data, plotting the phase map corresponding to this.
3. Repeating step 2 by taking different zernike polynomials and creating a data base which will be useful for the analysis of the real data. This will also prepare me to do the fitting models easily.
4. Collecting data from the IFO and applying these fitting models to it. Finding the set of zernike polynomials which are similar to the actual fugure error of the mirror. Plotting the Phase map corresponding to those zernike polynomials.
If you feel that there is some mistake in the steps, please correct me. It will be really helpful!
The values obtained from both analytical and finesse solution is given in the above table along with the corresponding percentage errors.finesse1.pdf
The parameters used for this calculation are listed below.
The cavity scan data obtained from Finesse is also attached here.
Aim: To develop a neural network in order to correlate the intensity fluctuations in the scattered light to the angular motion of the test mass. A block diagram of the technique employed is given in Attachment 1.
I have used Keras to implement supervised learning using neural network (NN). Initially I had developed a python code that converts a video (59 sec) of scattered light, after an excitation (sine wave of frequency 0.2 Hz) is applied to ETMX pitch, to image frames (of size 480*720) and stores the 2D pixel values of 1791 images frames captured into an hdf5 file. This array of shape (1791,36500) is given as an input to the neural network. I have tried to implement regular NN only, not convolution or recurrent NN. I have used sequential model in Keras to do this. I have tried with various number of dense layers and varied the number of nodes in each layer. I got test accuracy of approximately 7% using the following network. There are two dense layers, first one with 750 nodes with a dropout of 0.1 ( 10% of the nodes not used) and second one with 500 nodes. To add nonlinearity to the network, both the layers are given an activation function of tanh. The output layer has 1 node and expects an output of shape (1791,1). This model has been compiled with a loss function of categorical crossentropy, optimizer = RMSprop. We have used these since they have been mostly used in the image analysis examples. Then the model is trained against the dataset of mirror motion. This has been obtained by sampling the cosine wave fit to the mirror motion so that the shapes of the input and output of NN are consistent. I have used a batch size ( number of samples per gradient update) = 32 and epochs (number of times entire dataset passes through NN) = 20. However, using this we got an accuracy of only 7.6%.
I think that the above technique gives overfitting since dense layers use all the nodes during training apart from giving a dropout. Also, the beam spot moves in the video. So it may be necessary to use convolution NN to extract the information.
The video file can be accesses from this link https://drive.google.com/file/d/1VbXcPTfC9GH2ttZNWM7Lg0RqD7qiCZuA/view.
Gabriele told us that he had used the beam spot motion to train the neural network. Also he informed that GPUs are necessary for this. So we have to figure out a better way to train the network.
gautam noon 11Jun: This link explains why the straight-up fully connected NN architecture is ill-suited for the kind of application we have in mind. Discussing with Gabriele, he informed us that training on a GPU machine with 1000 images took a few hours. I'm not sure what the CPU/GPU scaling is for this application, but given that he trained for 10000 epochs, and we see that training for 20 epochs on Optimus already takes ~30minutes, seems like a futile exercise to keep trying on CPU machines.
I was checking on the slow machine channels and found something I could not understand.
On the IOO WFS HEAD screen, there are two sets of 4 switches (magenta rectangles in Attachment 1) labeled 2/4/8/16dB.
But as far as I could confirm with the WFS demod (D980233) and WFS head (D980012) drawings, they are the gain (attenuation) switches for the individual segments.
Their epics variable names are "C1:IOO-WFS1_SEG1_ATTEN", "C1:IOO-WFS1_SEG2_ATTEN", etc...
"C1:IOO-WFS1_SEG1_ATTEN", "C1:IOO-WFS1_SEG2_ATTEN", etc...
I confirmed the switches are alive (effective), and they are not all ON or OFF. I wonder what is the real situation there...
Unfortunately, this has happened (and seems like it will happen) enough times that I set up a script for rebooting the machine in a controlled way, hopefully it will negate the need to repeatedly go into the VEA and hard-reboot the machines. Script lives at /opt/rtcds/caltech/c1/scripts/cds/rebootC1LSC.sh. SVN committed. It worked well for me today. All applicable CDS indicator lights are now green again. Be aware that c1oaf will probably need to be restarted manually in order to make the DC light green. Also, this script won't help you if you try to unload a model on c1lsc and the FE crashes. It relies on c1lsc being ssh-able. The basic logic is:
Given the various changes to the IFO config since last Thursday when I was last able to lock the DRMI, I wanted to try once again tonight. However, I had no success. By my judgement, the alignment is fine as judged by looking at mode flashes on the cameras. However, despite following the usual alignment procedures, I did not get a single lock in tonight.
Perhaps we can use a flip mount on the BS that combines the PSL and AUX beams on the AS table, so we have the option of recovering the usual IFO config when we so desire - while Jon needs the SRC locked for his measurement, it would be nice to not have to figure out the correct demod phases etc each time there is a change in the optical setup of the AUX beam.
Among the things that we hadn't taken care of yesterday before beginning to look for transmission signals were the polarization of the AUX beam on the AS table and optimizing the PLL feedback. The AUX beam is s-polarized on the PSL table (choice due to availablility of mirrors), and I added a half waveplate in front of the fiber to match it's axes. I placed another half-waveplate at the fiber output and send the reflection port of a PBS cube onto a PDA1CS photodetector. By alternatingly turning the waveplates I minimized the reflected light, giving strongly p-polarized light on the AS table for best results when interfering with the IFO beam. I wiggled the fiber and found no strong dependency of the output polarization on fiber bending. Attachment 2 shows the current layout.
The beat signal between AUX and PSL table is at -20dBm, and I adjusted the PLL gain and PI-corner to get reliable locking behavior. I think it's a good idea to keep the AUX beam on the AS table blocked while it's not in use, and only unblock it when it is phaselocked to avoid a rogue beam with no fixed phase relation to the PSL in the IFO.I blocked the beam after completing this work today.
I used the signal chain that Keerthana, Koji, and I set up yesterday to look for mode flashed of the AUX light in the YARM using the RF beat with the PSL carrier in transmission. To align the AUX beam to the arm the following steps were performed:
This was followed by a sweep over two full FSRs. Attachment #1 shows the trace recorded by the AG4395 using the max data hold setting during the sweep. Essentially the beat between AUX and PSL carrier traced out the arm's transmission curve. At minimum transmission there was still a ~82dB beat on the transmission PD visible.
The YEND QPD is currently blocked and sees no light.
It isn't clear to me in the drawing where the Agilent is during this measurement. Over 40m of cabling, the loss of signal can be a few dB, and considering we don't have a whole lot of signal in the first place, it may be better to send the stronger RF signal (i.e. Marconi pickoff) over the long cable rather than the weak beat signal from the Transmission photodiode.
I worked a bit on the PSL table today