I've been banging my head against bilinear noise subtraction, and figured I needed to test things on some real hardware to see if what I'm doing makes sense.

I ran the ASS dither alignment on the Y arm, which ensures that the beam spots are centered on both mirrors. 

I then drove ITMY in yaw with some noise bandpassed from 30-40 Hz. It showed the expected bilinear upconversion that you expect from angular noise on a centered beam, which you can see from 60-80 Hz below

I looked at the length signal, as the noise subtraction target, and the ITMY oplev yaw signal plus the transmon QPD yaw signal as witnesses.

There is some linear coupling to length, which means the the centering isn't perfect, and the drive is maybe large enough to displace it off center. However, the important part is the upconverted noise which is present only in the length signal. The QPD and oplev signals show no increased noise from 60-80Hz above the reference traces where no drive is applied

I then compared the multicoherence of those two angular witnesses vs. the multicoherence of the two (linear) witnesses plus their (bilinear) product. Including the bilinear term clearly shows coherence, and thereby subtraction potential, at the upconverted noise hump. 

So, it looks like the way I'm generating the bilinear signals and calculating coherence in my code isn't totally crazy.

I realized I had overlooked an important constraint in the layout, which is that the enclosure will have two supports that occupy some region of the table - these are denoted in blue in v3 of the layout (Attachment #1). I measured the dimensions for these from the existing Y-endtable. The main subsystem this has affected is the IR transmission monitors, but I've been able to move the photodiodes a little to accommodate this constraint.

I've also done the mode-matching calculations explicitly for the proposed new layout (Attachments #2 and #3, code in Attachment #4). While the layout was largely adopted from what Andres posted in this elog, I found that some of the parameters he used in his a la mode code were probably incorrect (e.g. distance between the 750mm lens and the ETM). More critically, I think the Gouy phase for the optimized solution in the same elog is more like 60 degrees. I found that I could get a (calculated) Gouy phase difference between the two PZT mirrors of ~81 degrees by changing the green path slightly, and making the two PZT mirrors Y7 and Y8 (instead of Y7 and Y11, for which the Gouy phase difference is more like 50 degrees). But this way the two steering mirrors are much closer to each other than they were before. Other misc. remarks about the mode matching calculations:

• The beam diameter at the locations where the Faraday isolators should go is well below 5mm, the aperture size of the Faraday isolators
• The calculated mode-matching efficiencies suggest that we don't need any cylindrical lenses though the mode from the NPRO is elliptical
• Attachment #5 is a CAD drawing of the layout with all dimensions used for the mode-matching calculations included (although they are in inches)

These changes also necessitated minor changes to the transmitted IR beampath and the Oplev system, but these changes are minor. I've also switched the positions of the AUX IR power monitoring PD and the fiber coupler as suggested by Koji. The shutter has also been included. 

I'm planning to start removing components from the X endtable tomorrow morning at ~10AM - if anyone thinks I should hold off and do some further checks/planning, let me know before this so that I can do the needful.

Local earth quake 3.1 magnitude in Valencia, Ca did not trip our suspensions.

There is currently no table at the X end!

We have moved the vast majority of the optics to a temporary storage breadbord, and moved the end table itself to the workbench at the end. 

Steve says Transportation is coming at 1PM to put the new table in.

As discussed in a Wednesday meeting some time ago, we don't need to be writing channels from BLRMS filter modules to frames at 16k (we suspect this is leading to the frequent daqd crashes which were seen the last time we tried setting BLRMS up for all the suspensions). EricQ pointed out to me that there conveniently exists a library block that is much better suited to our purposes, called BLRMS_2k. I've replaced all the BLRMS library blocks in the sus_single_BLRMS library block that I made with there BLRMS_2k blocks. I need to check that the filters used by the BLRMS_2k block (which reside in /opt/rtcds/userapps/release/cds/common/src/BLRMSFILTER.c) are appropriate, after which we can give setting up BLRMS for all the suspensions a second try...

I've begun cleaning the optics that will eventually go back onto the newly installed X-endtable. We decided that First Contact was the way to go (as opposed to methanol drag wiping). Koji demonstrated the application of the (red) First Contact solution onto a 2" mirror - I then proceeded to work on the rest of the optics. We are broadly following the procedure in E1000079 - first one coat of First Contact solution is applied, then a small piece of PEEK is embedded by applying a second layer of solution over it (this will enable us to pull off the First Contact once we are ready - the plan is to do this after roughly placing the optic on the table. As of now, I've finished coating most of the optics that are part of the IR Transmon path - I will continue later in the evening.

The new endtable is almost ready for re-population. Steve just needs to shim the enclosure which will be done tomorrow morning. The game-plan as discussed at the meeting today is to first try and set up the IR Transmon path. This will allow us to verify that the endtable height is such that we can maintain a beam height of 4" everywhere on the table (I suspect we may have to compromise at some poing and do some fine adjustment of 1/4 to 1/2" somewhere though). It will also allow me to define the cavity axis relative to the table, which will be useful to place the green steering optics eventually. Doing this will be challenging though as right now, I can't see any of the arm flashes on the endtable using an IR card. Ideally, we want to somehow lock the X arm and then do the checks mentioned at the endtable, before beginning to put the endtable back together. 

Steve has finished installing the enclosure on the new endtable. So Eric and I decided to try and lock the X arm and measure the beam height of the transmitted IR beam relative to the endtable. We initially thought of using POX DC as a the LSC trigger but this did not work as there was no significant change in it when the arm was flashing. Eric then tried misaligning the ITM and using AS110 as a trigger - this worked. We then recompiled the ASS model to take AS110 as an input, and ran the dither alignment. After doing so, I measured the beam height at two points on the new endtable.

Bottom line:

• The beam is roughly level across the table (along the North-South direction, within the precision to which I could place the irides and measure the height). The table has also been levelled pretty well...
• The beam height is ~4.7" across the endtable

So the beam is about 0.7" higher relative to the endtable than we'd like it to be. What do we do about this?

• Is it even possible to raise the table by 0.7" so we can have a level beam everywhere? Are there some constraints related to how the enclosure is attached to the window?
• Are we okay with tolerating a solution where we keep the beam level at 4", and use Y10 and Y11 (see layout in elog 12060) to raise the beam by 0.7", and then have slightly higher posts for the optics downstream of this point?

I've also placed two irides extending the cavity axis on the endtable. These should be helpful in aligning the green to the arm eventually.

The new TMC 4' x 3' x4" optical table and enclosure is installed - aligned- leveled.

Atm2,  Picture is taken ~42" from the window at 3.75 camera height. The leveled table height is wthin 1/4 at the center of the window.

I think this is close enough to move on with the installation of the optics.

We can raise the loaded table in the future if it is needed.

Atm4, Optical table height to floor 33" at the south west corner

Atm3, Enclosure top cover transmission at 1064 nm, 1mm beam size, power level 157 mW, 0 degree incident angle,   T 1.3% Metal shield is required above 100 mW hitting the wall of the enclosure!

Atm5, window to enclosure Kapton seal

X arm resonating after alignment, beam height on ETMX optical table ~4.75"

I have copied over the complete frame files from two DRFPMI lock acquisitions + locks to /frames/archive. The data should be safe from the wiper script here.

One, under the subfolder DRFPMI_Mar29_cal is the lock where the CAL-DARM channel is properly calibrated at GPS time 1143274087.

The other lock, under DRFPMI_MAR29_nocal, does not have the calibration set up yet, but was a much quicker acquistion (<2 min from ALS acquisition to DRFPMI) and longer lock (~8min).

The laser is back. Test report is in the 40m wiki as New Pump Diode Mephisto 1000

It will go on the PSL table.

The ruby wire standoff V groove cuts are looking good.

I will request free sample of  sapphire prizm where one side would have SOS's R cylindrical surface.

The present plan to have the v-groove on this prism.

Did it again.

PMC Trans ~0.739
IMC Trans ~15000

Over the last couple of days, I've been working on restoring the optical layout on the X-endtable. Some notes about the status as of today:

Lightwave NPRO output power

The output power from the lightwave NPRO is about 210mW (as measured with the calorimeter). This is significantly lower than the value of ~300mW reported in this elog. It may be that the laser crystal temperature has changed compared to that measurement, but the "ADJ" parameter is at 0, both today and in that measurement. The laser has also been on for more than a day now, that should be sufficient time for the crystal to equilibriate to its final operating state? Is such a large change in output power possible just because of a change in laser crystal temperature? Or did the laser really lose ~1/3rd of its output power over the last two months?

Alignment into IR Faraday, and changes to the planned layout

I've set up the layout until steering the beam through the IR faraday. The input power into the IR Faraday is ~210mW. The output power is ~186mW, after optimizing the angle of the HWP. These numbers seem consistent with what I had reported in this elog (although this was for the Innolight NPRO). The alignment looks reasonably good to the eye as well.

I've made one change to the planned layout (latest version here). Y1 is now a 2" 99% reflective for S polarization beam splitter, instead of a 1" HR mirror. I made this change because we want some light from the NPRO to be transmitted through this optic to couple into the fiber eventually, for the IR beat. I measured the transmitted power to be ~1.5mW, which is around what we were coupling into the fiber before, and should suffice now. The Lightwave NPRO datasheet (page 4) suggests that the polarization of the output of the laser is S, and the measured power before and after this optic suggests that it is working as advertised. This means that HWP 1 also has to be moved downstream (to rotate the polarization so as to maximize transmission through the IR faraday). Space constraints meant that I could not mount HWP 1 on the baseplate+3/4" OD post assembly which is what we want where possible on the new table, so for this optic, I used a 1" OD post and a fork. There may be a couple of other optics in the final layout where space constraints dictate we compromise in this way.

I've also installed beam dumps for the rejected light from the Faraday. For now, these are the old beam dumps. They looked reasonably intact. I believe we have a bunch of new beam dumps on hand as well, so these can be swapped out if deemed necessary.

Cleaning of optics

All the optics are being cleaned using first contact before being installed on the table. 

As I found out the hard way, it is not a good idea to clean small optics like half-wave plates while in their mounts. The first contact tends to bond to the frame while drying, and doesn't come off cleanly. Koji helped me clean the offending pieces (he used tweezers to manually remove the residual first contact, and then some acetone to clean up any remaining residue). Subsequently, he re-cleaned these optics, again using first contact, but this time being careful not to extend all the way out to the edge of the optic. The idea is to cover as much area as possible with first contact, while staying clear of the edge. This approach worked reasonably well.

The next major step is to achieve optimal alignment into the doubler. I've placed the doubler on the table in it's approximate final position, I wanted to make sure the enclosure support wasn't in the way (it isn't). The cable from the oven won't run all the way to the Thorlabs temperature controller in it's usual place, we need to either extend the cable, or figure out a new place where we can keep the temperature controller.

ETMX optical table is grounded to ETMX chamber through 1 Mohms

The doubling oven temp controller is installed to reach its cable.

[Koji,gautam]

Lightwave NPRO information:

Model: 126-1064-700

Serial Number: 337

Manufactured: December 1998!!

Details of checks performed:

Koji tuned the parameters on the laser controller and we observed the following:

1. Turning "ADJ" to +10 and the pumping current all the way up to the maximum (2.62A) allowed us to recover an output power of 300mW, at a laser crystal temperature of ~45degrees
2. The output power increased almost monotonically as a function of the laser crystal temperature - why? We were able to see powers as high as 250mW (at ADJ=0) for the maximum crystal temperature of ~60 degrees.
3. We checked that we could believe the readout of the power meter by measuring the power using the Scientech power meter - we saw ~270mW after the Faraday with this meter, accounting for ~10% loss through the Faraday, this corresponds to an output power of 300mW (all this was done at ADJ=+10, DC=2.62A). I suspect that the display is dodgy though, because changing the Diode Current from 2.52A to 2.62A increased the output power by almost 100mW, which seems hard to believe?
4. The Lightwave NPRO does not have heat dissipation fins attached - could this be affecting the power output somehow? In any case, this has to be rectified. So if we decide to keep the Lightwave NPRO, the layout will still need minor changes to accommodate the heat fins. Steve, do we have these in hand?

Way forward

Ericq has begun the characterization of the repaired Innolight. We checked that it outputs 1W of power. We will now have to perform the following measurements:

• Frequency noise using PLL
• AM/PM response of the PZT
• Laser power output as a function of diode current - this will be useful for diagnostic purposes in the future
• AUX temperature vs PSL temperature at which beatnotes can be found
• Waist measurement - the mode matching and optical layout upstream of the doubling oven at least will have to be modified significantly

All of these will have to be done before installing this laser at the endtable.

I believe the consensus as of now is to go ahead with carrying out the above measurements. Meanwhile, we will keep the Lightwave NPRO on and see if there is some miraculous improvement. So the decision as to whether to use the Innolight is deferred for a day or two.

Just a heads up that some equipment is hooked up at the PSL table for the repaired AUX laser PLL measurement, I plan to continue with it tonight.

I've taken a few spectra that, along with the PZT coefficient from the repair sheet, that suggest the noise level is ok (incoherent sum of AUX and PSL at about ~3e4 / f Hz/rtHz), but calibrated plots, etc. will follow in time.

The free running PSL+AUX beat frequency noise spectrum has been measured via PLL. AUX laser PZT PM and AM responses were measured too. 

Rough notes about these measurements:

Laser -> QWP -> HWP -> PBS -> 10% BS -> Beat
3.4Vpp out of PD, (40% contrast)
20dB Coupler, output to analyzer, coupled output to Mixer (-a few dBm, didn't check specifically)
Mixer: ZP-3+, BLP-5.1 at output
LO: OCXO @ 36MHz 13dBm->5dB Att-> +8dBm LO at Mixer

Got ~65mVpp out of Mixer

Mixer out -> SR560, LP 3Hz, G=500 -> Pomona Summing node -> Laser PZT
~30kHz UGF ~30 deg phase

Spectra, OLG via SR785 taken with free running PSL, anthropomorphic temperature servo. Data sheet calibration used for PZT. SR560 output noise dominates over analyzer, mixer, PD. Spectrum looks ok, I think.

PM measured with AG4395. High impedance probe used for laser PZT, otherwise couldn't lock. PM calibrated via mixer voltage span for fringe-to-fringe. 

PSL beam blocked, AUX power increased to read 8.0V, AM measured with AG4395.

AM/PM doesn't look to dissimilar to old measurements on wiki. ~230kHz looks like a fine modulation freq. 

Still to be done to AUX laser:
- joint PSL/AUX temperature sweeps
- Output power vs. diode current
- Beam profile

I've performed the temperature sweep of PSL vs Innolight 1W AUX laser.

• I followed the procedure in this elog - started by turning of FSS and FSS Slow servos, closed the PSL shutter, noted down the value of PSL temperature
• As noted in elog 3759, there are multiple temperatures at which a beat can be found. I recorded all that I could find. The IR beat frequency was < 20MHz at the temperatures recorded (and had an amplitude of a few dBm, but I used a 20dB coupler to look at the signal on the HP spectrum analyzer
• The PMC unlocked each time I changed the PSL temperature, but the PMC autolocker worked for me every time
• We should use curve 3 in attachment 1, it is the most reliable set of temperatures at which a beat can be found
• PSL diode current was 2.100A, AUX laser diode current was 2.001A
• Attachment 2 is the data

It remains to measure the output power vs diode current, and the beam profile. I will do the latter on the SP table where there is a little more space. Because we have 1W from this NPRO, the knife-edge method requires a power meter that has a large dynamic range and is sensitive enough to profile the beam accurately. After consulting the datasheets of the power meters we have available (Scientech, Ophir and Coherent) together with Koji, I have concluded that the Coherent calorimeter will be suitable. Its datasheet claims it can accurately measure incident powers of up to 100uW, although I think the threshold is more like 5-10mW, but this should still be plenty to get sufficient resolution for a Gaussian intensity profile with peak intensity of 1W. We also checked that the maximum likely power density we are likely to have during the waist measurement process (1W in a beam of diameter 160um) is within the 6kW/cm^2 quoted on the datasheet.

I re-measured the power levels today.

We have ~205mW out of the NPRO, and ~190mW after the Faraday. It doesn't look like the situation is going to improve dramatically. I'm going to work on a revised layout with the Innolight as soon as I've profiled the beam from it, and hopefully, by Monday, we can decide that we are going ahead with using the Innolight.

I have moved the 1W Innolight + controller from the PSL table to the SP table for beam profiling.

Summary

I've finished up the remaining characterization of the repaired 1W Innolight NPRO - the beamscan yielded results that are consistent with an earlier beam-profiling and also the numbers in the datasheet. The output power vs diode current plot is mainly for diagnostic purposes in the future - so the plot itself doesn't signify anything, but I'm uploading the data here for future reference. The methodology and analysis framework for the beamscan is the same as was used here.

Attachment #1 - Beam-scan results for X-direction

Attachment #2 - Beam-scan results for Y-direction

Attachment #3 - Beam profile using fitted beam radii

Attachment #4 - Beam-scan data

Attachment #5 - Output power vs Injection current plot

Even though I remember operating at a diode current of 2.1A at some point in the past, while doing this scan, attempting to increase the current above 2.07A resulted in the "Clamp" LED on the front turning on. According to the manual, this means that the internal current limiting circuitry has kicked in. But I don't think this is a problem as we don't really even need 1W of output power. This is probably an indicator of the health of the diode as well?

Attachment #6 - Output power vs Injection current data

It remains to redo the mode-matching into the doubling oven and make slight modifications to the layout to accommodate the new laser + beam profile. 

I plan to do these in the morning tomorrow, and unless there are any objections, I will begin installing the repaired 1W Innolight Mephisto on the X endtable tomorrow (18 April 2016) afternoon. 

The PSL had one 1064 nm beam to be blocked around the north east side. The end enclosures are fine.

Summary of work done over the last two days

1. Lightwave NPRO + controller moved to PSL table
   • ​​The interlock is not connected to the controller
   • Controller is not powered
2. Innolight NPRO + controller installed at endtable
   • ​​​​Interlock has been connected
   • For initial alignment purposes, I'm running it at an injection current of 1.000A (~50mW of IR out of the NPRO)
   • Temperature of crystal set to 31.66 degrees in anticipation of operation in the nominal state
3. Laying out optics
   • ​​Given that the mode out of the NPRO is different from that from the Lightwave, the mode-matching had to be re-done
   • Attachment #1 shows the mode-matching solution being implemented
   • Current state - I've placed all the optics up to and including the doubling crystal + oven. Alignment through IR Faraday is pretty good, QWP+HWP angles optimized to maximize transmission through the Faraday (<10% loss). Oven has been hooked up to temperature controller, and is currently set to 36.3 degrees. Coarse alignment into doubling crystal done at lower power. Even with the low IR power, I am able to see some green. It remains to turn the injection current up and do the fine alignment + lens position tweaking to maximize the green power from the doubling crystal - with ~1W of power, assuming 2%/W SHG efficiency, we should be seeing 20 mW of green (which is probably way too much)

Immediate next steps:

1. Some optimization to be done with regards to beam dumps for rejected beam from IR Faraday. Also double check to make sure that the reflected beam from L1 doesn't go back directly to the laser (at the moment it doesn't, is there a standard way to do this? I was trying to have the lens as close to normal incidence as possible, but I may not have been entirely successful which is why the reflected beam is not going straight back at the moment).
2. Optimize mode-matching into the doubling crystal
3. Once the desired green mode is obtained, continue with the rest of the layout
4. Update CAD drawing to reflect new layout

It is worth wiping table top covers. Use isopropanol soaked lint free wipes.

I've made progress on the new layout up to the doubling oven. After doing the coarse alignment with the diode current to the NPRO at ~1A, I turned it back up to the nominal 2A. I then rotated the HWP before the IR Faraday such that only ~470mW of IR power is going into the doubler (the rest is being dumped on razor beam dumps). After tuning the alignment of the IR into the doubling oven using the steering mirror + 4 axis translation stage on which the doubling oven is mounted, I get ~3.2mW of green after the harmonic separator and a HR mirror for green. The mode looks pretty good to the eye (see attachment #1), and the conversion efficiency is ~1.45%/W - which is somewhat less than the expected 2%/W but in the ballpark. It may be that some fine tweaking of the alignment + polarization while monitoring the green power can improve the situation a little bit (I think it may go up to ~4mW, which would be pretty close to 2%/W conversion efficiency). The harmonic separator also seems to be reflecting quite a bit of green light along with IR (see attachment #2) - so I'm not sure how much of a correction that introduces to the conversion efficiency. 

While doing the alignment, I noticed that some amount of IR light is actually transmitted through the HR mirrors. With ~500mW of incident light at ~45 degrees, this transmitted light amounts to ~2mW. Turns out that this is also polarization dependant (see attachment #3) - for S polarized light, as at the first two steering mirrors after the NPRO, there is no transmitted light, while for P-polarized light, which is what we want for the doubling crystal, the amount transmitted is ~0.5%. The point is, I think the measured levels are consistent with the CVI datasheet. We just have to take care find all these stray beams and dump them.

I will try and optimize the amount of green power we can get out of the doubler a little more (but anyway 3mW should still be plenty for ALS). Once I'm happy with that, I will proceed with laying out the optics for mode-matching the green to the arm.

Our last RGA scan is from February 14, 2016  We had a power outage on the 15th

Gautom has not succeded  reseting it. The old c0rga computer looks dead. Q may resurrect it, if he can?

Dan sealed the leak today.

Earth quake stops need viton tips.

Wirestandoffs are still aluminum.

Gautom is progressing with the layout nicely. The X-arm transmission window have not seen cleaning for decades. This should be the time to do it. Here is picture of dirtiness.

It is not that simple... How much effort should we put in it? The hole table with 1W inno laser plus... set up now about ~500 lbs  We can pull it off carefully, but it is not risk free.

We should look at our other signal port windows! Gautom's long reach able him to do the first contact cleaning without moving anything. It is great!
 

Attachment #1

Layout as of today. Most of the green path is done. The Green REFL PD + PZT mirrors have not been hooked up to their respective power sources yet (I wonder if it's okay to start laying cables through the feedthroughs on either end of the table already, or if we want to put whatever it is that makes it airtight eventually in first?). A rough power budget has been included (with no harmonic separator just before the window), though some optimization can be done once the table is completely repopulated.

Attachment #2

A zoomed-in version of the REFL path.

Some general notes:

1. I've tried to use the custom 3/4" O.D. posts + baseplate arrangement wherever possible (only 1 steering mirror is on a 1" post clamped with a fork to the table because of space constraints). Where the baseplates could not be bolted onto the table directly, I've used Newport SS Dogs to do the job.
2. I checked for continuity between the PZT outer case and the table top with a multimeter, and found none. So I chose to leave the Thorlabs baseplates in place. For the REFL PD, I've used an insulating baseplate given to me by Steve.
3. I've used some custom length 3/4" O.D. posts to get the beam up to the right height (~4.75") just before sending the green beam in. The beam height is 4" elsewhere.
4. I was playing around with positioning the harmonic separator immediately before the vacuum chamber window - I found that there is a substantial amount of green light that is reflected, though there doesn't seem to be any IR leaking through. The mirror was labelled Y1-1037-45P, which is a code for CVI mirrors, though I believe it is a LaserOptik product and that we have a couple of other such mirrors in the optics cabinet - though they are all 1". This document suggests that from the back side, there should be <0.1% reflection of green while on the front side it should be < 3%. I will have to hunt a little more for the specs, and measure the powers to see if they match the previously quoted numbers. In any case, I'll have to think of how to separate the (unwanted) reflected green and the transmitted IR from the cavity in the IR transmon path.
5. There are some minor changes to the planned layout posted here - I will update these in due course once the Transmon path and Oplev have been set up.

I am closing the PSL shutter and the EX laser shutters for the night as I have applied a layer of first contact to the window for cleaning purposes, and we don't want any laser light incident on it. It may be that the window is so dirty that we may need multiple F.C. cleaning rounds, we will see how the window looks tomorrow...

                   Sad situation

    The anti-symmetric port

spider webs fly in the wind

Bah, we need ruby slippers for all future suspensions. Prism with curved backside and smooth grooves.

No aluminum, no cry.

It looks very promising.

The IR Transmon system is almost completely laid out, only the QPD remains to be installed. Some notes:

1. The "problem" with excessive green power reflected from the harmonic separator has been resolved. It is just very sensitive to the angle of incidence. In the present configuration, there is ~10uW of green power reflected from either side, which shouldn't be too worrisome. But this light needs to be dumped. Given the tiny amount, I think a black glass + sticky tape solution is best suited, given the space constraints. This does not reach the Transmon PDs because there is a filter in the path that is transmissive to IR only. 
2. I aligned the transmitted beam onto the Thorlabs PD, and reconnected the signal BNC cable (the existing cable wasn't long enough so I had to use a barrel connector and a short extension cable). I then reverted the LSC trigger for the X arm back to TRX DC and also recompiled c1ass to revert to TRX for the dither alignment. At the moment, both arms are stably locked, although the X arm transmission is saturated at ~0.7 after running the dither alignment. I'm not sure if this is just a normalization issue given the new beam path or if there is something else going on. Further investigations tomorrow.
3. It remains to dump some of the unwanted green light from the addition of the harmonic separator...
4. We may want to redesign some (or all) of the Transmon path - the lens currently in use seems to have been chosen arbitrarily. Moreover, it is quite stubbornly dirty, there are some markings which persist after repeated first contact cleaning...

I feel like once the above are resolved, the next step would be to PDH lock the green to the arm and see what sort of transmission we get on the PSL table. It may be the polarization or just alignment, but for some reason, the transmitted green light from the X arm is showing up at GTRY now (up to 0.5, which is the level we are used to when the Y arm has green locked!). So a rough plan of action:

1. Install transmon QPD
2. PDH lock green to X arm
3. Fix the window situation - as Steve mentioned in an earlier elog, the F.C. cleaning seems to have worked well, but a little remains stuck on the window (though away from where any laser radiation is incident). This is resolved easily enough if we apply one more layer of F.C., but the bottle-neck right now is we are out of PEEK which is what we use to remove the F.C. once dried. Steve thinks a fresh stock should be here in the next couple of days...
4. Once 3 is resolved, we can go ahead and install the Oplev.
5. Which leaves the lst subsystem, coupling to the fiber and a power monitor for the NPRO. I have resolved to do both these using the 1% transmitted beam after the beamsplitter immediately after the NPRO rather than pick off at the harmonic separator after the doubling oven. I need to do the mode-matching calculation for coupling into the fiber and also adjust the collimating lens...
6. Clean-up: make sure cables are tied down, strain-relieved and hooked up to whatever they are supposed to be hooked up to...

The 2F product out of the mixer is a natural concern when demodulating. However, I think this isn't so big of a deal in our green PDH servos; 420kHz isn't so high of a frequency that the servo amplifiers are bandwidth or slew-rate limited. Furthermore, the amplitude of this line is supressed by the loop somewhat, since it arises from the same field product that the loop is acting on. Measuring the Y end mixer output with a high impedance probe and the AG4395 shows it to be something like -50dBm. 

In fact, the main thing that the pomona LPFs are accomplishing right now is filtering the 1F content of the mixer output that arises from the second order sideband creating a signal at 2F, and beating with the LO at (2F-1F)=1F. This line is something like -30dBm (5mVrms) at the mixer output; I can reproduce this amplitude with a back-of-the envelope calculation using a modulation depth of 0.3, 8V out of the PD at DC when unlocked, the mixer datasheet, and the nominal cavity parameters. 

The nice thing about this is that we don't need to filter this after the mixer, we can use a [bandpass/lowpass/notch] filter before the mixer (as is done in the LSC demod boards) to filter out the 2F (420kHz) content of the PD signal, which will only introduce some small amount of linear time delay to the PDH loop, instead of the wicked phase loss from the current post-mixer LPF. We can then replace that 70kHz filter with something of lower order or higher corner frequency to win a good deal of phase in the PDH loop. 

OK - but give us a circuit diagram and the expected before/after loop plots. Got to make sure we keep the right impedance from PD to mixer. Some of the Thorlabs PDs have a 50 Ohm instead of 0 Ohm source impedance. Maybe you can try it out now since the green arm is ready.

Using the modulation frequency suggested here, I hooked up the PDH setup at the X-end and succeeded in locking the green to the X arm. I then rotated the HWP after the green Faraday to maximize TRX output, which after a cursory alignment optimization is ~0.2 (I believe we were used to seeing ~0.3 before the end laser went wonky). Obviously much optimization/characterization remains to be done. But for tonight, I am closing the PSL and EX laser shutters and applying first contact to the window once more courtesy more PEEK from Koji's lab in W Bridge. Once this is taken care of, I can install the Oplev tomorrow, and then set about optimizing various things in a systematic way.. MC autolocker has also been disabled...

Side note: for the IR Transmon QPD, we'd like a post that is ~0.75" taller given the difference in beam height from the arm cavity and on the endtable. I will put together a drawing for Steve tomorrow..

After a second round of F.C. application, I think the window is clean enough and there are no residual F.C. pieces anywhere near the central parts of the window (indeed I think we got most of it off). So I am going to go ahead and install the Oplev. 

We can get as much, if not more, attenuation of the 1F line in the mixer output that we get from the post-mixer LPF from using the following passive filter between the PD and mixer RF input:

There should still be some kind of LPF after the mixer, but I haven't yet determined what it should be; this will determine how much phase the PDH loop wins. At most, this should win around 25 degrees at 10kHz.

The filter was designed by referencing the "Handbook of Filter Synthesis" by Zverev, looking for an elliptic filter for matched source and load impedences, 40dB min attenuation in the stopband, a stopband frequency that starts at twice the corner frequency, and minimizing the VSWR between the PD and filter in the passband.

In terms of the tables in the book, this means: n=5, rho=2%, theta=30deg, K**2 = 1.0. The dimensionless component values were scaled by the corner frequency of 200kHz, and reference impedence of 50 Ohm. (The corner is a little lower than the real modulation frequency, since the nonzero resistance of the inductors pushes the frequency up a bit)

The ideal capactior values do not correspond to things we have in hand, so I checked our stock and chose the closest value to each one.Unsurprisingly, due to these component substitutions, and the fact that the coilcraft inductors have a resistance of about 7 Ohms, the predicted TF of the realizable filter does not match the design filter exactly. However, the predicition still looks like it will meet the requirement of 40dB of supression of the 2F line in the PD signal. (Since we have tunable inductors, I've used the ideal inductor values in generating the TF. In practice I'll inspect the TF while I tune them)

[In this TF plot, I've multiplied the real response by 2 to account for the voltage division that occurs with ideal 50 Ohm impedance matching, to make 0dB the reference for proper matching]

The filter's phase delay at the modulation frequency is just about 180, which as a time delay of 5usec works out to 9 degrees of phase loss at 10kHz in the PDH loop. According to some old measurements, the current LPF costs something like 35 degrees at 10k, so this wins at most around 25 degrees, depedent on what LPF we put after the mixer.

LISO source both traces is attached!

Antonio/Gautam are now developing a more up to date Finesse model of our recycling cavities to see what we can have there before our power recycling gain or cavity geometric stability is compromised. Expect that we will here a progress report on the model on Wednesday.

Some thoughts:

1. RC folding mirrors need to be dichroic to allow green beams to get out.
2. We should look at the specs Jamie used to get the RC folding mirrors last time and figure out what went wrong / what specs to change.
3. T_1064 < 100 ppm. Hopefully < 50 ppm.
4. On the AR side, we mainly want low AR for green, but nothing special for 1064, since that's taken care of by the HR.
5. How much should we wedge these things?
6. Should the wedge be horizontal?
7. Can we get someone in Downs to update the optical layout?
8. What microroughness do we need?
9. The mirrors must be flat, with the  500 m < RoC < 100 km. Part of the Finesse modeling is to figure out what happens if the RoC is in the 300 - 1000 m range. Better stability?

With Steve's help, I installed the Oplev earlier today. I adjusted the positions of the two lenses until I deemed the spot size on the QPD satisfactory by eye. As a quick check, I verified using the DTT template that the UGF is ~5Hz for both pitch and yaw. There is ~300uW of power incident on the QPD (out of ~2mW from the HeNe). In terms of ADC counts, this is ~13,000 counts which is about what we had prior to taking the endtable apart. There are a couple of spots from reflections off the black glass plate in the vacuum chamber, but in general, I think the overall setup is acceptable.

This completes the bulk of the optical layout. The only bits remaining are to couple the IR into the fiber and to install a power monitoring PD. Pictures to follow shortly. 

Now that the layout is complete, it remains to optimize various things. My immediate plan is to do the following:

1. Maximize green transmission by tweaking alignment. I should also do a quick check using mirror specs to see that the measured transmitted green power compares favourably to what is expected.
2. Check the green PDH loop transfer function at the X end - this will allow me to set the gain on the uPDH box systematically.
3. Re-establish green beats, check noise performance.
4. There are possibly multiple beam dumps that have to be installed. For now, I've made sure that no high power IR beams are incident on the enclosure. But there are a couple of red and green beams that have to be accounted for.

I will also need to upload the layout drawing to reflect the layout finally implemented.

Not directly related:

The ETMx oplev servo is now on. I then wanted to see if I could lock both arms to IR. I've managed to do this successfully - BUT I think there is something wrong with the X arm dither alignment servo. By manually tweaking the alignment sliders on the IFOalign MEDM screen, I can get the IR transmission up to ~0.95. But when I run the dither, it drives the transmission back down to ~0.6, where it plateaus. I will need to investigate further. 

GV Edit: There was some confusion while aligning the Oplev input beam as to how the wedge of the ETM is oriented. We believe the wedge is horizontal, but its orientation (i.e. thicker side on the right or left?) was still ambiguous. I've made a roughly-to-scale sketch (attachment #1) of what I think is the correct orientation - which turns out to be in the opposite sense of the schematic pinned up in the office area.. Does this make sense? Is there some schematic/drawing where the wedge orientation is explicitly indicated? My search of the elog/wiki did not yield any..

[ericQ, gautam]

Today we spent some time looking into the PDH situation at the X end. A summary of our findings.

1. There is something that I don't understand with regards to the modulation signal being sent to the laser PZT via the sum+HPF pomona box - it used to be that with 2Vpp signal from the function generator, we got ~5mVpp signal at the PZT, which with the old specs resulted in a modulation of ~0.12rad. Now, however, I found that there was a need to place a 20dB attenuator after the splitter from the function generator in order to realize a modulation depth of ~0.25 (which is what we aim for, measured by locking to the TEM00 modes of the carrier and sidebands and comparing the ratio of powers). It could be that the PZT capacitance has changed dramatically after the repair. Nevertheless, I still cant reconcile the numbers. We measured the transfer function from the LO input of the pomona box to the output with the PZT connected, and figure there should be ~70dB of attentuation (with the 20dB additional attenuator in place). But this means 1Vpp*0.0003*70rad/V = 0.02rad which is an order of magnitude away from what the ratio of powers suggest. Maybe the measurement technique was not valid. In any case, this setup appears to work, and I'm also able to send +7dBm to the mixer which is what it wants (function generator output is 3Vpp).
2. In addition to the above, I found that the demodulated error signal had a peak-to-peak of a few volts. But the PDH servo is designed to have tens of mV at the input. Hence, it was necessary to turn down the gain of the REFL PD to 10dB and add a 20dB attenuator between mixer output and servo input.
3. While Johannes and I were investigating this earlier in the afternoon, we found that the waveform going to the laser PZT was weirdly distorted (still kind of sinusoidal in shape, but more rounded, I will put up a picture shortly). This may not be the biggest problem, but perhaps there is a better way to pipe the LO signal to the PZT and mixer than what is currently done.
4. We then looked at loop transfer function and spectrum of the control signal. Plots to follow. They look okay.
5. I measured the green power coming onto the PSL table. It is ~400uW. After optimizing alignment, the green transmission is ~0.4 according to whatever old normalization we are using.
6. We then recovered the X green beatnote and looked at the ALS noise spectrum. Beatnote amplitude at the beat PD is ~ -27dBm. The coherence in the region of a few hundred Hz suggests that some improvements can be made to the PDH situation (the gain of the PDH servo is maxed out at the X end at the moment...). But the bottom line is this is probably good enough to get back to locking...

We took an OLG measurement of the green PDH loop. It seems consistent with past measurements. I've added a trace for the the post-mixer lowpass, to show its contribution to the phase loss. (EDIT: updated with measured LPF TF)

I used this measured OLG and the datasheet laser PZT conversion factor to calibrate the control signal monitor into the AUX laser frequency noise, it looks consistent with the frequency noise measured via the PSL PLL (300 Hz/rtHz @ 100Hz). Above a few tens of kHz, the control signal measurement is all analyzer noise floor, due to the fourth order 70kHz lowpass after the mixer (the peaks change height significantly depending on the analyzer input range, so I don't think they're on the laser). Gautam will follow up with more detailed measurements of both the error and control signals as he noisebudgets, this was just intended as a quick consistency check.

This morning I poked around with the green layout a bit. I found that the iris immediately preceding the viewport was clipping the ingoing green beam too much, opening it up allowed for better coupling to the arm. I also tweaked the positions of the mode matching lenses and did some alignment, and have since been able to achieve GTRX values of around 0.5.

I also removed the 20db attenuator after the mixer, and turned the servo gain way down and was able to lock easily. I then adjusted the gain while measuring the CLG, and set it where the maximum gain peaking was 6dB, which worked out to be a UGF of around 8kHz. On the input monitor, the PDH horn-to-horn voltage going into the VGA is 2.44V, which shouldn't saturate the G=4 preamp stage of the AD8336, which seems ok.

The ALS sensitivity is now approaching the good nominal state:

There remains some things to be done, including comprehensive dumping of all beams at the end table (especially the reflections off of the viewport) and the new filters to replace the current post-mixer LPF, but things look pretty good.

All seems very fishy. Its not good to put attenuators and filters in nilly-willy.

1. Once the post-PD bandpass has been designed and constructed, you should be able to use whatever PD gain setting gives you the best SNR. There's no need to use more PD gain than necessary; it just reduces the PD bandwidth. What is the input referred current noise of the PD at the different gain settings?
2. The open loop mixer output *should* be very large. It should be reduced to mV only when the loop is closed.
3. The better way to estimate the modulation depth is to lock the arm on red as usual and then scan the EX laser and look at the green transmission. The FSR is 3.7 MHz, so the SBs should show up well in a narrow scan around the carrier.
4. I guess its going to be tough to impedance match the splitter box to the NPRO PZT, since its impedance is all over the place at 200-300 kHz, but you could put a 50 Ohm in-line terminator in there somewhere?
5. The Bode plot seems to indicate that we could easily get a 10 kHz UGF and then switch on a Boost. Is the remote Boost switch disabled or always ON? I am suspicious of the plot and think that the coarse trace is probably missing some sharp resonances which will sneakily bite you.

I suggested in an earlier elog that after the repair of the NPRO, the PZT capacitance may have changed dramatically. This seems unlikely - I measured the PZT capacitance with the BK Precision LCR meter and found it to be 2.62 nF, which is in excellent agreement with the numbers from elogs 3640 and 4354 - but this makes me wonder how the old setup ever worked. If the PZT capacitance were indeed that value, then for the Pomona box design in elog 4354, and assuming the PM at ~216kHz which was the old modulation frequency was ~30rad/V as suggested by the data in this elog, we would have had a modulation depth of 0.75 if the Function Generator were set to output a Signal at 2Vpp (2Vpp * 0.5 * 0.05 * 30rad/V = 1.5rad pp)! Am I missing something here?

Instead of using an attenuator, we could instead change the capacitor in the pomona box from 47pF mica to 5pF mica to realize a modulation depth of ~0.2 at the new modulation frequency of 231.25 kHz. In any case, as elog 4354 suggests, the phase introduced by this high-pass filter is non-zero at the modulation frequency, so we may also want to install an all-pass filter which will allow us to control the demodulation phase. This should be easy enough to implement with an Op27 and passive components we have in hand...

I've build the filter, and it seems to have the desired TF shape.

I also re-purposed the 70k lowass to a ~120k lowpass by changing the 68nF caps to 22nF caps, since we still want some post-mixer rolloff. 

However, putting the ELPF in the chain caused some weird shapes in the OLG. I still need to get to the bottom of it. However, just with the post-mixer LPF modification, here's what the OLG looks like:

As Rana surmises, we definitely still add a boost and maintain a 10k UGF. I still need to look into the state of the remote boost....