Contact with Coherent re Diabolo

I got in contact with Coherent they have provided some instructions on aligning the laser cavity, something I've already attempted and knew how to do.  For future reference I've attached the instructions on the wiki (they are confidential): they can be found attached HERE.

More details on Diabolo manual and documentation can be found on the Manuals-And-Datasheets ATF wiki page.  

Temperature tuning

I've been attempting to find sweet spot for phase matching temperature between oven and self heating.  The FWHM of the phase matching temperature of the SHG lithium niobate crystal appears to be on the order of about 0.2 C which makes it quite narrow.  Given there is quite a bit of self heating in the cavity when it is locked up, it would be very easy to miss the correct temperature when setting the crystal's oven set point.

Koji/rana has suggested turning down the laser power, locking and then gradually tuning the oven temperature to get to ideal set point as we increase power.  I've attempted this and found that the autolock requires some minimum amount of laser power to engage.  I opened up the controller box and traced a few things through the PCBs, but it is not immediately apparent which pot adjusts the autotune threshold value for power.  I've emailed the contact at Coherent and they are making inquires to see if they can tell me what to adjust to lower this threshold manually.  For now I will attempt a few more full power locks at carefully spaced values of crystal temperature to see if I am able to happen upon a sweet spot.

The strategy locking at lower laser power and optimizing phase matching temperature by slowly walking the temperature and laser power up won't work: the autolocker mode of the Diabolo SHG won't engage at lower laser power. Instead I tried a brute force approach of stepping temperature and re-locking the cavity in 0.05 C increments of set point temperature.  I've plotted the result below.  Error bars are based on standard deviation on 32 averaged measurements over 10 seconds.

I stopped at 106.00 C as  power output was not improving beyond that point. Maximum temperature is clipped at the controller at 110.26 C. Peak output of 66 mW is actually  at 110.20 C. This is counterintuitively higher than the previous set point but might be good enough to work with, as long as its stable over longer periods.

Still not clear what has changed to shift the ideal SHG over set point to 1.4 C higher than it was before.  Maybe the SHG oven needs to go even a little hotter, but we can only access tempertures up to 110.26 C. I'll take what I can get.

Some other laser settings: 

• Injection current: 2.077 A
• Laser crystal temperature: 26.32 C
• SHG PDH gain: 0.5
• PDH offset: 4.95
• Noise eater: off

Data and plotting notebook are attached in a zip below: 20181220_SHGLockedPowerVsSetPointTemp.zip

This means that you want to make the SHG crystal longer. Is that true? If so, can you change the temperature for the optimal phase matching by tuning the 1064 crystal temperrature? I suspect you need to cool the YAG crystal, but I am not sure what is the thero-optic constant of the SHG crystal, and how much you can gain from this.

No I don't think so.  The temperature of the oven is important for the phase matching condition.  Its the nob you can turn so that the refractive index for both 532 nm and 1064 nm is just right so that both traveling waves stay in phase as they propagate in the non-linear crystal.  Otherwise the phase precesses in one wavelength relative to the other causing repeated amplification and de-amplification as the light propagates (rather than just amplification when they are in phase).  

Its true that temperature change induces expansion​ and dn/dt changes that will change the round trip effective length/phase. However, the PZT in the SHG should pick up the slack of from changes in the crystal round trip phase due to temperature.  We can tune the laser temperature to bring the PZT to the center of its range but at this stage it doesn't look like its railing the range of this actuator.  

My guess is that self heating affects the relative alignment of the crystal HR surface relative to the PZT mounted (curved) mirror, this small change in alignment will shift the eigen axis of the cavity and mean that the locked hot cavity will have a slightly different optimal alignment to the unlocked case.  Its much harder to walk alignment of 1064 nm while keeping cavity locked.

I think I've managed to bring the power of the Diabolo back to a usable level for WOPO.

More power with misalignment

I had another look at the alignment into the SHG.  This time I systematically trialed a range of cavity misalignments to deliberately set the pointing into the cavity to be wrong for the fast scanning mode of the control unit (cold) but to allow for the locked (hot) cavity to expand into correct alignment.  This video shows the SHG output as the cavity is locked IMG_4766.MOV,  this seems to show the expansion affecting mostly the horizontal axis. However, eventually I found that the locked cavity had a preferred misalignment on the vertical axis (in terms of the nobs that I needed to turn).  In reality its a mixture of horizontal and vertical once all the DOF are traced through the lenses and mirrors.

I was able to get a steady 160 mW of power with this new misalignment/alignment strategy.  The photos below shows the 1064 nm transmission peaks and error signal for a combination of (mostly) vertical and (a little) horizontal misalignment​ that gave greater power. 

The oven operating temperature was at 110.14 C with a laser diode current of 2.102 A.  PDH gain was set to 0.56 and offset was set to 5.1.

Upping the oven temperature limit

The Diabolo SHG oven set point temperature range doesn't put the ideal phase matching temperature at the center of the available values.  In fact, the optimal power for the above-mentioned power boost was achieved pretty much of the edge of the available range (top of range is 110.26).  I popped the lid on the controller box and traced through all the PCBs to find out how the temperature range was trimmed. The front panel potentiometer (1k) is trimmed with a small set screw (blue) potentiometer located directly to the low left of the temperature knob (when viewed from the component mounting side).  I trimmed the set point temperature value so that the upper value was 111.00 C, giving an extra 0.75 C of headroom above the ideal phase matching set point.  

I found that at the very upper edge of the new range the thermal control loop started to loose some stability.  I limited the increase of temperature to the 111.00 C point and will use some caution when adjusting temperatures at these upper ranges to make sure I don't get oscillations in power.

Great recovery job!

I had another look at MM the 532 nm light into the P3-460B-FC-2 patch cable.  After walking some lens positions and mirror pointing based on ATF:2257 I found that I could get 1.07 mW output from launch of 2.3 mW (46.5 %) through the fiber.  I expect the most I'll need is about 30 mW at the output which will put the amount of required power at about 64 mW.  There is now enough 532 nm laser light to do this and I think its within the tolerances of the fiber to have that much launched in.

One thing I did notice, on closer inspection of the fiber ends, is that there is a little bit of damage on one end (labeled A) of the P3-460B-FC-2 patch.  This is pictured below.

I tried a bit of gentle cleaning with some fiber cleaning cloth (Thorlabs FCC-7020)  but this appears to be a burn mark when zoomed in. Sorry doesn't capture very clear on my phone through the fiber microscope. I swapped the end that was in the fiber collimator (F240APC-532) but didn't increase the fiber launch efficiency​.  Not sure when this got damaged, but I might have exposed it to excessive power at some stage. 

I still think that the fiber through put should be sufficient, if it turns out to be a problem then re-ording shouldn't be an issue.  Thorlabs usually seems to make these fiber items next day.

Next steps

Next step is to tune up the 1064 nm alignment into the LO launch fiber and see if I can get the new PZT mounted mirror to scan as expected.  Need to figure out if there is a way to calibrate the phase on this so that I can check it is actually scanning phase (the last PZT broke and I had no idea).

Need to figure out how much 1064 nm light can be co-propagated in the patch waveguide.  The previous dismal power values might be because I made very little effort to mode match,  I'll look at what lenses I have available and see if I can boost the power through put to something respectable.

Alignment of the pumping 532 nm polarization into the WOPO is important to getting the correct phase matching condition.  For the periodically polled Lithium Niobate (LN) waveguide the phase matching is type-0: and pumping and fundamental wavelengths are in the same polarization.  The AdvR non-linear device is coupled with polarization maintaining fibers (Panda style), which are keyed at their FC/APC ends.  This means that with the correct launch polarization we should be correctly aligned with the proper crystal axis for degenerate down conversion (at the right chip temperature). 

Replacing Broadband PBS

Till now I was using non-pol maintaining patches to coupling into the WOPO fiber ends.  This should have been ok, but it is hard to figure out exactly which polarization is optimal so I switched to a pol-maintaining patch because it can be aligned separately and then the keyed connectors give you automatic alignment.  I had some issues trying to find the optimal polarization going into the fiber and I've now traced this back to the polarizing beam cubes.  I've been using Thorlabs PBS101 which is a 10x10x10 mm^3 beam cube that is supposed to be broad-band (420-680 nm).  When I checked the extinction ratio I saw Pmax=150 mW, Pmin=0.413 mW on transmission between extremes.  This is an extinction ratio of Tp:Ts = 393:1 which is much less than the spec of >1000:1.  Not sure what's going on here, the light going into the BS is coming directly from a Faraday isolator and a half-wave plate.  With some adjustment to the angle of the wave plate I can do a little better but it should be nicely linearly polarized to start with.

I've switched out the PSB101 for the laser line PBS12-1064 I remeasured extinction ratio (Pmax=150 mW, Pmin=27.6 µW) Tp:Ts = 5471:1 (better than the quoted 3000:1 spec).  This is good, at least now I know what is going on. I am also putting in an order for a 532 nm zero order quarter-wave plate, so that we can be absolutely sure we are launching in linear light always.  

Aligning light into pol-maintaining fiber

I previously thought I might be able to use the frequency modulation technique to align the light through the polarization maintaining fiber.  There is a birefringence in PM460-HP fiber of  3.5 x 10^-4.  The phase between ordinary and extraordinary axes over the whole fiber length is

Where L is fiber length,  is the birefringence and f is the laser frequency.  The idea is to launch linearly polarized light into the fiber and then at the readout place a polarizer rotated to be 90°: ramping frequency will produce an amplitude modulation on the dark fringe.  However, even with 1 GHz of frequency ramp this is only a 15 mrad effect for a 2 m fiber, its likely to be too small to see over other effects.  This is not enough to be able to fine align polarization.  

Instead I'll use the heat gun method.  I'll fire linearly polarized light into the fiber and measure the output with a crossed polarizer.  If the input polarization is correct there should be no power changes on the output as the fiber is thermally cycled. Its only two meters long so hopefully this effect is easy to see.

I want to get signal from about 1 MHz down to around DC from my subtracted  homodyne photodetectors. I'm planning to do something like this:

This mixer has a 4.78 dB conversion loss and should do the job.  Only issue is that to operate the mixing down the low pass filter needs to be lowered from 1.9 MHz down to something pretty low to ensure the harmonic (2 MHz term is removed).  These are 4th order filters, we'd probably want the cut off to be an order of ten below the mixing frequency... 100 kHz.  I don't see this in the minicircuits catalog don't know how doable that is to make one.  I'll have a look at what the 40 m has.

Making 5th order Elliptical Filter

I couldn't find any filters that would cut off above 100 kHz so I made my own using a Thorlabs EEAPCB1 generic filter PCB in a Thorlabs EEA14 enclosure. I used a 5th order elliptic design with a pass band up to 100 kHz and a stop band of 40 dB from 150 kHz.  To speed things up I used the Coilcraft Filter Designer software  (4.0.1, Windows) and chose closest standard values from parts we had in EE workshop kits. The Coilcraft designer is nice because it has the full physical model of the inductors built in.

The schematic is illustrated below:

Actual values selected were closest available and I didn't try to do any mixing or matching to get fine tuned correct values. Values are as follows:

• L2 and L4 27 µH => 1812CS-273XJL (27µH @ 2.5 MHz, tol 5%, Q_min 15 @ 2.5 MHz, SRF_min 11 MHz, DCR 11.04 Ω)
• C1 20.47 nF => 12065C223KAT2A (0.022 µF, 50 V, 10%)
• C2 14.38 nF => 12065153KAT2A (0.015 µF, 50 V, 10%)
• C3 47.42 nF => 12061C473KAT2A (0.047 µF, 50 V, 10%)
• C4 4.792 nF => 12065C472KAT2A (4700 pF, 50 V, 10%)
• C5 27.3 nF => 12065C273KAT2A (0.027µF, 50 V, 10%)

The built filter is shown below.

The transfer function was taken from 10 Hz up to 5 MHz (IF BW of 10 Hz).  The Coilcraft Filter Designer software seems to export all the filter scattering parameters except S12 and S21, not the best. Instead I used LISO to model the filter's predicted response and this is plotted alongside measured TF.  Here I have scaled the filter model's response by 2 to match the impedance condition of the TF measurement.  The coilcraft 0805LS-273X_E 27 µH inductors series resistance was modeled as 11 Ω (as per spec sheet) and values of capacitors were those used in the actual circuit.  In the original ideal 5th order elliptical circuit there is a double dip above the corner frequency.  The series resistance of the non-ideal inductors dampens these.  I don't really want to spend much more time on mixing and matching capacitor values.  It looks like for now that the pass band ripple is acceptable and the attenuation is >40 dB at 1 MHz and 2 MHz where we are trying to block harmonic signals. I'll leave optimization of this part for now and write this off as done.

Edit Mon Feb 25 19:22:14 2019 (awade): Fixed the phase pane of the bode plot, had accidentally used magnitude and wrong units. Also fixed some spelling.

I've set up a rotating PBS and half-wave plate to provide polarization adjustment into the 532 nm fiber without misalignment the spatial alignment.  Here I've used a PRM1 rotation mount with a SM1PM10 lens tube mount for beam cube prisms.  The lens tube mount is supposed to be for pre-mounted cubes but I've inserted some shims to hold it in place and it seems to work well like that.  It means I can get a nice clean linear polarization at all rotations.

After spatially aligning the input beam I stepped the rotation of the PBS (and accordingly the L/2 wave plate) and pulsed the temperature of the fiber using a heat gun.  After some walking I found that for the current fiber rotation (0 deg) the linear polarization was aligned with the fiber axis at 88 deg PBS rotation (here 0 deg PBS rotation is aligned for p-pol transmission, well almost). I made some adjustments to the alignment of the fiber collimator in the fiber launch, I aligned the slow axis key with the vertical so that the fast axis of the fiber is p-pol.

Keying on PM fibers

As a side note the keying of PM fiber patches is typically with the slow axis aligned with the key notch. The WOPO's PM fibers are keyed so that the alignment key is along the slow axis of the fiber (i.e. aligned with the stress rods). Figure below illustrates the configuration.

Replacing the 532 nm patch with fresh PM fiber

I was getting a large jitter in the power levels as measured at the output of the old SM and PM fibers (on the order of 10%).  These power fluctuations were not present on the input side.  I thought this was an alignment jitter or a polarization effect.  However, I was unable to minimize it by improving the input polarization at the launch.  When I tapped various mounts there didn't seem to be a corresponding correlation with output power jitter of the fiber.  When I checked the end of the PM fiber (P3-1064PM-FC-2), I saw that there was damage about the core (see pictured below).  It seems like maybe I had some kind of etalon effect from this burn mark and the launch.  After replacing the 532 nm PM fiber with a fresh one that arrived last week the power is much more stable and I was able to easily​ find the pol alignment going in. 

Next job is to replace SM fiber for the 1064 nm delivery with PM fiber so there is a well defined polarization for launching into the homodyne detector.

I've replaced the SM fiber in the 1064 nm launch with a PM fiber (P3-1064PM-FC-5). I also moved the fiber collimator (F240APC-1064) back 2.54 cm back to give more space for a PBS cube (to check linearly of the light).  

For the 1064 nm launch it seemed to be a lot harder to find the initial alignment of the collimator using the alignment of the back propagated 650 nm fiber laser source. Here I aligned a pair of irises in the forward propagating direction and then back propagated through the PM fiber using 650 nm to get the initial​ pointing of collimator. I don't know why this is so much harder than the 532 nm case.  I suspect one of the steering mirrors is not really reflecting off the front dielectric surface.  In the end I did a bunch of systematic walking of the fiber launch mount and eventually fount the alignment.  

From 4.44 mW of input light I get 2.74 mW of light out the other end of the fiber.  This is an efficiency of 62 % which is more than enough for my needs.  I expect the HD will only need 1 mW (2 mW max), so this is fine. Getting this in coupling higher will require a bit of lens walking, not really worth it at this stage.

I had already carefully aligned the collimator orientation to put the fast axis on aligned to p-pol (wrt the table), by eye.  It seems like the launch pretty much hit the correct launch polarization on the first go.  I see little variation in the polarization when I pulse the heat on the fiber.  This is now good to go for optimizing the homodyne visible and polarization overlap output from the SQZ.

These are things to get do this week on WOPO experiment:

☑️ Reinstall 50:50 fiber splitter into homodyne setup and go back to fiber launch of both ends of the HD directly onto photodetectors (rather than free space)

☐ Check visibility of HD by launching 1064 nm into both arms of HD using splitter and extra length on one arm, ramp laser frequency to get fringes and look at Vpp on each diode seperatly

☑️ Optimize subtraction of HD for max Common Mode Rejection (CMR) of LO amplitude noise.  Inject 3.21 kHz line into laser BNC port and minimize this peak on the subtracted output of the HD.

☐ Check 1064 nm -> 532 nm conversion in WOPO device to establish polarization basis for correct pol alignment into fiber (change 532 nm launch polarization if necessary), should be along the fast axis but for some reason this isn't in the datasheet explictly

☐ Double check how much power change there is from 2 pi modulation depth from PZT mounted 1064 nm mirror.  We don't want to over actuate on this element as it slightly misaignes into the fiber launch and causes some change in power, this is ok over a small range as the CMR of the homodyne will reject this.  We just want to be sure that this isn't the dominant effect that we are seeing at the output once we thing we should be seeing SQZ.

☑️Tick mark when done.

[awade, anchal ]

After a bit of reading I've realized that the standard use of these PM fibers is to launch along the slow axis (see for example Thorlabs and OzOptics info on fiber beam splitters).  It should be much of the sameness for patch cables, but polarization sensitive elements like beam splitters are mostly tested and specified for slow axis launch unless they are custom made to order. 

We are switching the polarization alignment to slow axis in the 1064 nm and 532 nm fiber coupling.  Anchal is re-optimizing​ the 1064 nm launch to get the PM fiber extinction ratio back to a good place.  We've also changed input launch to use a laser line PBS mounted in a rotation mount for clean linear polarization.  With the optimized setup the for the 1064 nm fiber path the output polarization signal goes from 3700 mV to 39.3 mV which is an extinction​ ratio of -19.7 dB.

Here the max theoretical extinction​ ratio is 

which would place our goodness of alignment to with 0.61 deg.

This is just a note about damage tolerances for fibers so we have a reference of the amount of power that can be used.  

Thorlabs gives the maximum theoretical CW power as 1 MW/cm² and a 'practical' safe power level of 250 kW/cm² (or a quarter of the max).  They don't seem to provide information about wavelength dependence and assume its the same for all wavelengths.  We don't expect to be affected by any of the exotic ultra high power effects like bend loss induced damage and photodarkening.  The air/glass interface is where the damage will occur.  This can either be because of heating of the ferrule/connector (causing epoxies etc to break down and damage the interface by depositing on the optical surface) or regular mechanism that are the same as bulk optics (dielectric break down and thermal effects).  

The intensity profile of light confined in the fiber is defined by the Mode Field Diameter (MFD) -- the cross-sectional diameter of the light that includes the core of the fiber and a region just beyond the cladding the mode occupies. MFD of 1064 PM fiber (PM980-XP) is 6.6 ± 0.5 μm @ 980 nm and for the 532 nm PM fiber (PM460-HP) is 3.3 ± 0.5 µm @ 515 nm.  

Fiber effective area is 

which is 8.5x10^-8 cm^2 for PM460-HP and 3.4x10^-7 cm^2 for PM980-XP.  Taking the conservative 'practical' damage threshold this indicates a maximum power of 21.4 mW into the 532 nm fiber and 85.5 mW into the 1064 nm PM fiber. The absolute maximum is just a factor of four more than this: 85.6 mW into 532 nm PM fiber and 342 mW into 1064 nm fiber. If the fiber ends are kept clean then we should be fine if the power level is kept below 85 mW.

Fiber beam splitter

Of particular concern is the power handling capability of the fiber beam splitter (PN1064R5A2).  There will be some waste 532 nm light coming out of the WOPO and I don't want these potentially multi-mode components to exit the cladding at the point of the coupler and damage the surrounding material. The 1064 nm maximum power rating of this 50:50 PM beam splitter is listed as 1 W (for the connectorized fiber), so we should be well clear of that threshold for the LO light.  For 532 nm its less clear. The equivalent 532 nm PM 50:50 beam splitter (PN530R5A2) has a rated power of 100 mW @ 530 nm for bare or connectorized fibers. As the MFD of the 1064 nm version of this PM beam splitter has a much larger MFD and the exiting 532 nm light will already be expanded in the cable patching the WOPO to the BS, we should be well clear of this damage threshold point.  

So bottom line is that we need to keep power below 85 mW going into the WOPO device and keep all the end connectors super clean and it will be fine.

Initial measurment of PD TF

I realize I never really measured the signal transfer function for each of these Photo Detectors (PD).  This post summarizes measurements of the optical to electrical signal output signal transfer function. Here I used the Jenne rig at the 40m and took a transfer function using an Aglent 4395A from 30 kHz to 5 MHz.  

I've labeled the two PD in the homodyne as unit A and unit B.  Below (first attached)  is the raw TF as measured by the Aglient taking the ratio of the detector to the reference NF1611 detector:

Fixing differences between output impedance unit A and B

After looking a bit at the circuit I realized that I had added 100 Ω in series in detector B (to limit current draw when driving 50 Ω loads).  This had not been added to the detector A, which means that with more power than 300 µW on this detector the op amp would be drawing more than 30 mA when driving a 50 Ω.  This shouldn't have been an issue when driving the summing circuit, but is good practice.  

I added 100 Ω in series with op27 output (thin film 1206 size) in detector A to match the output to that of detector B.  I remeasured the TF and get a much better match between the paths.

Calibrating TIA Measurment

The following is used to calibrate the  transfer function (See PSL:2247):

where Z_AC,PD is Calculated RF Transimpedance, Z _AC,Ref is the known RF transimpedance of the reference photodiode, 𝛿ϕ is arbitrary phase delay due to light and cable length and R_PC  is the photocurrent ratio at DC of reference PD to RFPD under test given by

For these measurements the calibration factors were as follows:

Unit A TF (after adding 100 Ω series​ output): measured NF1611 DC voltage of 780 mV (@ 1 MΩ impedance), PD DC output voltage was -1.20 V (@ 1 MΩ impedance).  Here the photodetector has transresistance is 6.8 kΩ. The NF1611 detectors have a DC path gain of 10 kΩ and an AC path gain of 700 Ω. From this the R_PC is 0.4496.  Here I ignore the overall phase (at ~1 MHz as the pi phase shift length is 150 m, so negligible).  

Unit B TF: measured NF1611 DC voltage of 791 mV (@ 1 MΩ impedance), PD DC output voltage was -1.18 V (@ 1 MΩ impedance).  Here the photodetector has transresistance is 6.8 kΩ. The NF1611 detectors have a DC path gain of 10 kΩ and an AC path gain of 700 Ω. From this the R_PC is 0.4558.  Here I ignore the overall phase (at ~1 MHz as the pi phase shift length is 150 m, so negligible).  

Results of calibrated transimpedance measurements are plotted below*. I also added the LISO estimated noise curve.  The 3 dB of unit A and B were 1.93 MHz and 1.72 MHz respectively.  There is some discrepancy with the LISO model here, some of this might be artifacts in the reference detector and some is likely dirt effects in the proto board circuit that I'm not going to debug for now.  The biggest discrepancy between the two detectors starts around 900 kHz.  When I looked again at the detectors I realized I had used a Wima foil cap in detector B for the bias LPF. It should be fine for these frequencies, but I switched the foil cap out and replaced it with a 1 µF + 100 nF ceramic cap to match the unit A.  I haven't gotten a chance to remeasure this since as Gautam is still using the Jenne rig to do some stuff. I'll remeasure later, we should then find the two responses match.

*I didn't do the many IRIS measurement​ uncertainty analysis, its not necessary at this stage. 

This set of measurments turned out to be bad. The OP27 (±15V supplies) with 50 Ω series at the output + 50 Ω load impedance seemed to be saturated.

The rest of this post is a stream of conciseness.

As noted in previous post (ATF:2316) the PD unit B in the homodyne had a slightly reduced HF response, possibly due to the use of a foil cap in the PD bias supply low pass.  I removed this and resmeasured.

Unit B TF: measured NF1611 DC voltage of  312 mV (@ 50Ω impedance), PD DC output voltage was - 566 mV (@ 50 Ω impedance).  Here the photodetector has transresistance is 6.8 kΩ. The NF1611 detectors have a DC path gain of 10 kΩ and an AC path gain of 700 Ω. From this the R_PC is xxx.  Here I ignore the overall phase (at ~1 MHz as the pi phase shift length is 150 m, so negligible).  This was a bad measurment, something was up with the circuit saturating way too low.  

Changed 100 Ω series on output to 50 Ω to make it more strait forward to understand how it works with 50 Ω load. 

Unit A TF: measured NF1611 DC voltage of  658 mV (@ 50Ω impedance), PD DC output voltage was 1.42 mV (@ 50 Ω impedance).  Here the photodetector has transresistance of 6.8 kΩ. The NF1611 detectors have a DC path gain of 10 kΩ and an AC path gain of 700 Ω. From this the R_PC is 0.3151.  Here I ignore the overall phase (at ~1 MHz as the pi phase shift length is 150 m, so negligible).  This measurment could have be saturating the current output of the OP27, likely that numbers here are not right.

After examining the output of detector B it became apparent that the output was saturated for powers of more than ~20 mA laser diode drive. 

Unit B TF: measured NF1611 DC voltage of  312 mV (@ 50Ω impedance), PD DC output voltage was -566 mV (@ 50 Ω impedance).  Here the photodetector has transresistance is 6.8 kΩ. The NF1611 detectors have a DC path gain of 10 kΩ and an AC path gain of 700 Ω. From this the R_PC is -0.3748.  Here I ignore the overall phase (at ~1 MHz as the pi phase shift length is 150 m, so negligible).   100 Ω  (50Ω + 50 Ω) loading is too low, OP27 open loop gain and voltage swing range slew starts to sag below 150Ω loading.  These results are bogus.

This time I measured with 50 Ω on the oscilloscope for apples-to-apples comparison with what the Aglient 4395A was seeing.  Turns out I think I was saturating the photodetector current wise in the first round of measurments. All of the above is probably not right and should be disregarded.  Also, later I switched to a AD829 as the op amp of choice so the above is not much use for characterizing the detector performance anymore.  The OP27 isn't really spec'ed to operate with loads this low and only has its full band with for output loads of >2 kΩ.  

[awade]

This post details the rebuild I made of the transimpedance amplifiers (TIA) for the homodyne detectors.  My previous iteration of this design failed to take into account the output swing limitations and current draw limitations at the output of the op amp.  It turns out that the OP27 was not the best choice here.  On the face of it its GBP and input referred noise were fine for a ~ 1 MHz detection.  However, operating the chip with 5V output voltage (to get 10 dB shot noise clearance from thermal noise) and with output impedance on order of 150 Ω meant that the frequency response was not as expected (see ATF:2316). 

A better choice is something like AD829 which is designed for driving signals directly into 50Ω terminated outputs.  The updated design is illustrated below.

Here the transimpedance gain has been lowered to 2 kΩ to lower the relative input referred noise contributed by the AD829's slightly higher current noise.  Here the optical power can be increased to give the same effective output signal.  In this case increase optical power and lowering gain keeps the thermal noise clearance below shot noise constant but improves the clearance of the AD829 input current noise.

Scaling of thermal noise relative to shot noise given DC output voltage

For reference the thermal noise of the TIA is given by the Johnson noise

where k_B is the  Boltzmann's constant, T is the temperature (300K) and Rfb is the feedback resistor value.  Also here the shot noise is

Where e is the charge on an electron (1.602e-19 C), R is the photodiode responsivity (~0.8 A/W), and P_PD is the power on the photodetectors (0.5-5 mW).  Keeping in mind that there is a maximum voltage swing (V_max) on the output of the op amp, this comes from its current driving capability into a given load, then the feedback resistance (TIA gain) is limited to 

The noise clearance, given a maximum output voltage, is then given by the ratio

Power on the photodiode and responsivity cancel out for the noise clearance.  Assuming we are not optical power limited then the thermal noise ratio (clearance below shot noise) is wholly determined by V_Max and the physical constants e, k_B and T.  For room temperature (300 K) this reduces to the rule of thumb 

Keeping the V_max to within 3 V means that we will get a thermal noise clearance factor of 7.6 below shot noise (8.8 dB). V_max of 5.2 volts will give 10 dB clearance. Ok I guess. 

Choosing AD829

Koji suggested the AD829 as a replacement to OP27.  AD829 has applications in driving 50 Ω/75 Ω loads in video applications and has some nice noise characteristics.  The bottom line is that it has 1.7 nV/√Hz input voltage noise, 1.5 pA/√Hz input current noise, can do a ±3 V voltage swing into 150 Ω load (DC coupled), and is fast (600 MHz uncompensated, with 230 V/µs slew).  There are some quirks.  AD829 seems have have a weird 80 MHz feature that causes oscillations if not compensated properly.  I did a bit of modeling in LISO and then just decided to build it once I found that ~2 kΩ gain was about right for ensuring that the dark noise wasn't dominated by the op amp current noise.

Because I was building on proto-board I wanted to ensure that there was as little parasitic capacitance as possible.  I built the whole amplifier + PD onto a single SOIC-14 to 0.1 pitch adaptor (Adafruit SMT Breakout PCB for SOIC-14, Part No. 1210), pictured below.  The small footprint and short trace lengths between critical pins means that there is a lower chance of parasitic impedance causing instability or impact on bandwidth.  

The feedback resistor (2 kΩ) and feedback compensating capacitor (15 pF ceramic + 3 pF nominal pin parasitic) were soldered directly to the reverse side where the TSSOP-14 pins were connected through vias to the top side.  This minimized the path length of this electrical signal.  I scratched off pads that were unused in case they shorted or were a source of crosstalk.  The photodiode mount pin was soldered directly to the edge of the board to minimize distance.  The power bypassing capacitors were SMT 100 nF ceramics (5%) that were soldered directly to the SMT breakout with the shortest path to ground on that board (see pictured).  The other photodiode pin was soldered to +5 V (-5V) supply pin in unit A (unit B).  The opposite biasing makes subtracting the signals using a summing circuit easier but may affect the overall TF response. There is no power regulation, I'm planning to use batteries to power these circuits.

In my initial tests I had no compensating capacitor attached to pin 5.  I inserted a 40 pF ceramic capacitor in place of the photodiode and looked at the AD829 output pin with a high impedance probe on an oscilloscope.  I immediately saw 80 MHz oscillations.  I don't really need a super high bandwidth so I went strait to the maximum recommended choice for pin5 compensating capacitor of 68 pF.  This makes AD829 unity gain stable with 66 MHz bandwidth (slew 16V/µs) but is more than enough for my needs.  This killed the high frequency RF ringing junk. Maybe less capacitance here would have worked, but this performance is enough for my needs and gives some certainty about the op amps stability (ignoring input capacitance compensation).

At the output I added 100 Ω of series resistance to limit the loading on the op amp and current draw when 50 Ω terminated.  With 50 Ω terminating impedance this makes a 150 Ω to ground.  Providing the output DC swing is kept within 3V the op amp should behave as expected.

The whole thing fits together very nicely on a single SMT breakout board. Making it compact will hopefully avoid any issues with stray capacitance messing up the performance.  I have then mounted this on a larger proto board for ease of installing in the experiment. Paths to the output SMA output are thin wire and not routed through the underlying board. Only power and ground is routed through the board, all other pins to the op amp disconnected. 

Transfer function PD

I measured the signal transfer function of the above TIA units using the Jenne rig at the 40 m.  It took me a number of tries: I made the mistake of loading the DC port of the NF 1611 detector with 50 ohms (which resulted in too much current draw and affected the LF response of the witness detector); there were also some issues with the polarization/positioning onto the beam splitter there that made it not 50:50 (it was 69:33 ), I moved the fiber launch a bit to get the splitting right and retook measurements after that; also I initially had the current set too high on the laser current driver which I think was saturating its response.  I'll skip over the details and just present the final measurement. I've included a zip in the attachments with the data that logs some more of the details of my failed attempts.

I used the same calibration as used in PSL:2247.  Here the laser diode current was set to 25.0 mA, which is 1.51 mW of 1064 nm, and the AC excitation into the laser was set to -21 dBm (19.9 mV rms).  Power on the witness detector was measured to be 0.79 mW and that on the PD under test was 0.75 mW (as measured with a power meter).  Units of measurement were linear magnitude and phase in degrees. DC voltages were measured with 1 MΩ impedance.

Unit A: using 25 mA current into laser and -21 dBm sweep I get DC on PD of 1.20 V (into 1 MΩ) and a DC voltage of NF1611 detector of -2.13 V (into 1 MΩ).

Unit B: using 25 mA current into laser and -21 dBm sweep I get DC on PD of 1.07 V (into 1 MΩ) and a DC voltage of NF1611 detector of -2.12 V (into 1 MΩ). 

The calibrated TIA current to voltage gain is plotted below:

A notebook detailing all the attempts is attached below.

Andrew measured the output voltage noise of photodiodes through an AC coupled pre-amplifier (G=200, 0.7 nV/rtHz) while varying the intensity of white light falling on it. The setup uses a regular 2xAA flashlight (torch) with an incandescent bulb to illuminate the photodiode. The flashlight focus is adjusted to give a non-expanding beam and a 25.4 mm (1") lens is then used to focus the collimated beam down. The total power of light was controlled by trimming it with an iris.

Data Analysis:

Below is the fitting of data to estimate the shotnoise intercept current. Here the fitted function is the quadrature sum of input referred shot noise and dark noise modeled as an equivalent shot noise quanitiy i_snint (called intercept current):

I have included zip file with data and notebook in it as well. With measured shotnoise intercept current of 48 ± 7 uA and 31 ± 7 uA, we are good to go to use these for measuring shot noise with light power in the order of 5.0 mW falling on the photodiodes. With 5.0 mW (~3.9 mA) on each PD, this gives us a clearance of 10dB from the dark noise floor on average for each detector.  Plots with fits are attached below.

As I couldn't find this on the 40m elog/wiki I'm putting notes here on how to configure a fresh webserver + python server for the Zurich box (model HF2LI) that runs an autobooting service job. 

The Zurich box (model HF2LI) has no front panel interface.  All controls are done from a USB interface to a host computer (running windows or Ubuntu/Debian).  This computer then hosts a web-server that runs an interface in a browser (through http) or responds to requests from python/matlab/labview/C API over the network.  

Installing data and webserver 

Set up on a Linux box is fast and easy from a fresh install.  A thin low powered computer is ideal. Start with a fresh install of Debian, use the latest stable release. Then download the latest stable release for the HF2LI, HF2IS, HF2PLL instruments from the Zurich downloads page.  You’ll probably want the 64bit linux. You can untar this file with

> tar xzvf LabOneLinux<arch>-<release>-<revision>.tar.gz

(Insert your filename details above)

Then cd into the directory it created and run

> sudo bash install.sh

It will ask a bunch of questions, just hit enter and y all the way through. The install should immediately boot the HF2 data server at the end.  You can check the data server is running with

> ziService status

Now plug in your Zurich box using the USB cable. Note: there are ethernet ports on the box, but these are NOT for network ethernet.  If the box is powered up the data server should find the box.  

Running the web server

In another terminal you’ll want to to start a Webserver so you can view the box status, settings and live data.  Go to another terminal and run

> ziWebServer

Your Zurich device should now be exposed on port 8006 of that machine.  You are read to go.

Now on that computer you can got to http://localhost:8006, alternatively on any other computer on the network you can go to the host machine’s IP address and see your Zurich interface.  I.e. if the computer IP is 192.168.1.121 then go to http://192.168.1.128:8006.  

Remote access

If you wanted to access the Zurich interface remotely (outside network) you can do some SSH forwarding if you have a gateway machine to outside the lab.  The standard command for this is 

> ssh -nNT -L <clientport>:localhost:<hostport> <controls>@<Host-IP-Address>

where controls is the username, Host-IP-Address is the IP address of the machine to be forwarded from, clientport is the port on your machine (the one you are ssh'ing from) and hostport is the port on the remote machine (for Zurich this is 8006).  Then when you go to http://localhost:<clientport> and you'll see Zurich interface.

Python API

Since the release 2.18 (Nov 2018) the python access tools are now just in Pypi.  You can install them with pip:

> pip install --upgrade zhinst

This should now allow you to use python API to access your Zurich box.

Setting up service to keep host alive from when box is booted

You’ll want to set up a service on your Debian box with systemd.  Make a unit

> cd /etc/systemd/system

> sudo vim ziServer.service

You must use sudo to be able to save the file here.

Paste the following in, this assumes that your username and its group is called controls on your machine:

Tip: in vim you can type “:set paste” to get exact paste without auto indent etc.

Save the file and exit.  You'll also need to create the working director, or change the above to your desired directory

> mkdir /home/controls/services/

Finally, tell systemd to look for new things:

> systemctl daemon-reload

and enable the service,

> systemctl enable ziServer.service

and manually start the service,

> systemctl start myservice.service

It should be running, try to see if this is so:

> ps | grep ziServer

It should run on reboot of the machine and restart every 30 seconds if it crashes for whatever reason. Try rebooting the host machine to check it is still running

Configuring web-server service

Follow the above steps to also auto load the webserver with a unit file named /etc/systemd/system/ziWebServer.service:

 Now you have a local network ready server for accessing the Zurich box through a web browser.

The detectors have been verified to have a low intercept current, that will give us good clearance of our shot noise to dark noise (at least as measured in the <100 kHz band).  The question is if we can make a ~ 1 Mhz measurement of squeezing without extra noise in that readout chain dominating over our expected 1 dB squeezing.

The plan is to use the Zurich box (model HF2LI) to mix down quantum enhancement from 1 MHz in our homodyne detector.  This will be well above most of the classical noise of the laser and other acoustic pickup from the environment.  

The Zurich box has a input referred noise of 5 nV/rtHz above 10 kHz and operating at at the higher end of its sampling rate. It has a couple of nice features that would make it well suited to the task.  We can use the lock in amplifier instrument to digitally mix down signal from 1 MHz and output one or both of the i/q quadratures to its digital output ports.  From this instrument we can also set the bandwidth of the 

Output noise of Zurich is 25 nV/rtHz, so I'll need some gain factor to boost signal if readout is done from the external ports. Relative gain of the two paths can be changed manually by typing values into boxes, but might be tought to do a direct ballenced subtraction within the box live if the real time drift is large.

Laser diode modulation port on Diabolo has a signal bandwidth of 5 kHz, so can't really be used to generate a reference signal.  

The Zurich box can also be accessed via a python API.  Its a little bit opaque in some points on how this works as their documentation is a little thin on actual examples.  I've included a jupyter notebook and equivalent .py file for future reference.  This code is tested and has worked with the Zurich box data server at IP address 10.0.1.23.  It shows exactly how settings can be modified and how data can be read back using their poll command.

Notebook includes all instructions for setting the Zurich HF2LI box up from scratch. 

For the detection of squeezed light the homodyne detector  is now configured for digital subtraction in post processing. The measurement scheme is now to take the output of the two TIA amplifiers (see QIL:2324 and QIL:2327), digitize​ them directly at 210 MSa/s (5 nV/rtHz input ref noise) with the Zurich box and demodulate both detector streams using its internal FPGA at about 1 MHz.  The Zurich box allows for direct sampling of the signal out of the FPGA which can then be downloaded either through the web interface or the python API. The signal chain is illustrated below:

The basic approach is to mix down the shot noise limited light from some higher frequency (somewhere in the range 100kHz - 8 MHz) and then compute the ideal subtraction factor from a short (10 sec) segment of data.  The method for optimization is to compute the time domain subtracted signal for a given balancing factor (SF) and compute the goodness of subtraction from the RMS of the signal.  An fmin search is done to find the ideal subtraction factor (SF) for shot noise limited light.

It wasn't immediately clear to me what demodulation​ frequency to choose, the low pass filter and the digitization sampling rate. So I set up a python notebook and did a few sweeps of these parameters to find the best combination of parameters. This book is attached below as 20190514notebook_ZurichDemodeCompairTests.ipynb in a zip. I've pickled data for replotting if needed later.  

Sweeping the demodulator 4th order LP corner frequency

At present the detectors are illuminated with 4.5 mA worth of light (as measured from DC voltages / TI gain).  This should give about 25 nV/rtHz at the output of the photodetector, as measured by the Zurich box.  Fixing the demodulation frequency at 500 kHz and the sampling rate at 30 kHz I found empirically that the when the LP filter (4th order) was kept at 10 kHz and below the minimum subtracted signal converged to about 30 nV/rtHz (about what I expected).  With the demodulation frequency set much higher than this, the minimum subtracted signal no longer converged to the predicted SN level.  Maybe there is some aliasing thing going on here.  The outcome of this experiment is that clearly the demodulation LP filter should be set to at least 1/3 of the sampling rate for the time domain rms minimization to work.  

Selection of sampling frequency

I did a similar sweep with the sampling rate while keeping the demodulation frequency fixed at 500 kHz and the demodulation low pass fixed at 10 kHz.  The conclusion was basically the same: as long as the demodulator low pass was kept within 1/3 of the sampling frequency then the method of computing the ideal subtraction factor using the total time domain rms as a cost function returned the minimum possible differential noise.

Selection of demodulation frequency

The demodulation frequency choice is a tricker one.  Initially I selected 1 MHz but was getting excess noise above what I expected for shot noise. I then looked at a single channel on the Zurich's scope in FFT mode.  Data is plotted below

Basically the plot shows that there is a strong peak at about 900 kHz. This is due to the relaxation oscillation of the laser.  When the laser's noise eater is turned on this peak is damped with some tradeoff of slightly higher noise at lower frequencies. So when choosing a demodulation frequency it is best to select something well below 1 MHz. I'm not sure if it is best to turn the noise eater off if the demodulation frequency is set well clear of the 900 kHz peak.  It seems like there is a noise penalty for the noise eater loop at all other frequencies. 

With the noise eater on I then did a sweep of the demodulation frequency while keeping the low pass filter at 10 kHz and the sampling rate at 30 kSa/s.  I've plotted the median ASD value in the band between 100 Hz and 1 kHz as an indication of the minimum SN level after digital subtraction verses the demodulation frequency.  Plot below.  There is a hump at about 1 MHz that should be avoided, and spikes at 87 kHz and just below 300 kHz but anything else in the range of 100 kHz to 700 kHz should be fine.  

Current selected configuration

So for now, based on the above sweeps of parameters, I have selected demodulation frequency of 600 kHz that is low passed at 10 kHz with a 4th order filter and sampled at 30 kSa/s.  As an example of optimized subtraction I have plotted the two homodyne channels along with the subtracted differential signal (optimized by minimizing subtracted time domain rms) and the dark noise (using the same subtraction value).

Here the clearance from the dark noise floor doesn't seem so great.  The input referred noise of the Zuirch box is 5 nV/rtHz per channel.  Also the equivalent output noise from the PDs at this point is 2.5 nV/rtHz.  As these are uncorrelated noise processes this would place the noise floor at about sqrt(5^2*2+2.5^2*2) = 7.9 nv/rtHz, pretty close to what we are actually seeing. 

One option here is to pre-amplify the PDs before going into the zurich box.  A standard mini-circuits amplifier like ZFL-500LN+ has an input referred noise of 1.6 nV/rtHz which would bring the dark noise floor down to something closer to 4.2 nV/rtHz.  I'll keep going without any pre-pre-amplification for now and see what I can do with 532 nm pumping.  There should be enough clearance to see at least something if it is there.

In order to see squeezing I want to scan the phase of the homodyne relative to the squeezed light and see the variations in noise power as a function of time. 

Excess 532 nm dumped on detector A

From my initial scans it seems like there is excess noise of a factor of 1.5 above shot noise (see below).  This is only present with 532 nm pump injected. It could be anit-squeezing washing out with some phase noise bluring across the sample time of 0.1 seconds or maybe RIN from residual 532 nm present at one of the photo detectors.

After injecting light into the WOPO for some initial tests it was apparent that there was about 1.4 mW of waste 532 nm light exiting the fiber launch on the detector A path. About 16 µW gets through to the detector from the dichroic mirror reflection. A quick measurement of 532 nm power on detector B show that there was about 49 µW coming out of that fiber end with 0.2 µW making to the photodiode. The 50:50 splitting doesn't apply for non-design spec wavelengths of fiber splitter.   The InGaAs detectors have a pretty poor responsivity at this wavelength but the pumping light on detector A was enough to create a DC voltage of 1.36 mV. After dividing through by detector gain of 2kΩ this is equivalent to about 0.68 µA of DC power on the detector A.  This suggests a responsivity of order 0.014 A/W.  There isn't enough light on detector B to create any DC voltage. 

The imbalance here, with the 532 nm light, means that there is a mechanism for coupling in 532 nm RIN into the measured output signal.  Not sure if the RIN would be all that high at the 600 kHz (that I'm planning to mix down from) but it would be a good idea to remove it anyway.  I'll look for a another HR1064/HT532 dicroic mirror to attenuate this 532 nm component a little more.

There are a few other possibilities for the excess noise when injecting pump, this is a checklist for me to run through tomorrow:

1. Its possible that the waste pump light from the 532 nm is heating the fiber beam splitter a bit.  This might cause imbalance when I go to measure the noise level for pumped WOPO.  I checked both channels detector on the oscilloscope with 532 nm turned on and off.  I couldn't discern a change in the balancing​.  I will need to look closer tomorrow with the actual zurich demodulated data.  Task for tomorrow is to compare the BS balancing with 532 nm on and off.  This can be done by tuning the WOPO temperature down by 15 C (to put it outside the phase matching region) and running the SN measurement with 532 nm on and off;
2. The excess wasted 532 nm my also be locally heating the SPDC chip, so the ideal phase matching temperature might not be right.  I will try a series of temperature steps much smaller than the FWHM of the phase matching sinc curve and see if turning the temperature down/up makes a difference;
3. Check the scanning of the 532 nm PZT is actually working.  Loop the patch cable back to the launch of the light into the fiber and use some of the dumped light from the 532 nm power control to make an interference measurement.  Here a large fringe visibility is not super important.  Just need to be able to count fringes/volt;
4. Need to verify how much light is making it into the WOPO.  This is difficult to do directly.  But a quick check is to see how much is making it out the other side.  The PM 1064 nm fiber should carry the 532 nm pump light out the other side of the WOPO without too much loss.  Check with power meter how much 532 nm is coming out other end of WOPO;
5. Need to document the level of power fluctuations of the 1064 nm LO light in the homodyne.  This could be leading to the fluctuating power level that is present after upping the demodulation band width and boxcar window.  If this is an issue I need to figure out quickly how to up the noise band width so that the boxcar window can be made very small and data capture is faster that the power varations of the 1064 nm.  If we had some DC monitor we might be able to correct this out in post processing;
6. I could just be smearing anti squeezing over both measurement quadratures.  Need to check my measurement sweep times and confirm my boxcar window for scanning noise as a function of time is narrow enough;
7. Consider if adding some minicircuits pre-amps before zurich would make measurment of SN -> SQZ clearance clearer.

I had a look at a few of these things.  I've found that it doesn't seem to be caused by heating of the beam splitter, while turning 532 nm on an off I see no change in the balancing of the beam splitter.  It appears that on remeasuring the data on Tuesday I found that there had been some glitching of the digitized data readout.  

Suspicions for now is that there is not as much pumping light making it to the chip as a thought and that bandwidth resolution of the time scan scan was a little too wide.  I have shortened the scan, increased the noise bandwidth of the demodulation and will widen the width of the low pass filter on the noise-time scan to increase the time resolution.

I'm also implementing proper subtraction of the signal that uses both the I and Q quadratures so that all the information about the relative phase and amplitude of the digitized signals is properly mixed down by the FPGA inside the Zurich.

Notebook with demo of the subtraction is attached in a zip below.

{Shruti, Yehonathan}

On Friday, we came down to QIL to poke around the WOPO setup. The first thing we noticed is that the setup on the wiki page is obsolete and in reality, the 532nm light is coming directly from the Diablo module.

There were no laser goggles for 532nm so we turned on the 1064nm (Mephisto) only. The pump diode current was ramped to 1A. We put a power meter in front of Mephisto and opened the shutter. Rotating the HWP we got 39mW. We dialed it back so that 5mW is coming out of the polarizer.

The beam block was removed. We disconnected the LO fiber end from the fiber BS - there is light coming out! we connected a power meter to the fiber end using an FC/PC Fiber Adapter Plate. The power read 0.7mW. By aligning the beam into the LO fiber we got up to 3.3mW.

We connected the BHD PDs to the scope on the table to observe the subtraction signal. Channel 2 was negative so we looked at the sum channel.

Time ran out. We ramped down the diode current and turned off Mephisto.

Next time we should figure out the dark current of the BHD and work toward observing the shot noise of the LO.

[shruti, yehonathan]

SHG and 532 nm beam alignment

Yehonathan brought over 532nm/1064nm laser goggles from the 40m.

• We turned on the 1064 nm Mephisto, and set the pump current initially to 1.5 A (which gave us ~30 mW of output)
• We then turned on the doubling crystal. It took a while to reach its setpoint temperature of 110 C. Initially (without adjusting the SHG cavity parameters) we saw less than a microW of power after removing the beam block and opening the shutter.
• Playing around with the SHG cavity settings, we realized that
  • Setting the switch to "Auto" is how to nominally operate the 532 nm beam
  • "Scan" is useful for scanning the cavity, the amplitude of which can be controlled by the "Scan amplitude" knob. "Standby" seems to effectively turn off second harmonic generation
  • "Offset" knob tends to change the amount of green power generated and adjusting the "Gain" simultaneously helps stabilize this power (with some difficulty)
• On increasing the laser driver current to 2 A and adjusting the temperature setpoint to 109.7 C, we were able to see 100 mW of green power!
• Next, we played with the alignment into the fiber, seeing that initially there was barely any power at the other end of that patch cable
  • We adjusted the waveplate near the laser head to give us 5.3 mW of power right before coupling into the fiber
  • Adjusting a single mirror (the nearer one) did not result in much gain, so we simultaneously adjusted the two steering mirrors and achieved about a 50% coupling. (Our readings suggested it could be the max)
  • At  the end of the alignment: 2.74 mW at the output and 5.46 at the fiber input
• Reconnecting the fiber to the waveguide, without adjusting the temperature of the advr waveguide, we saw that the fiber at the output of this crystal seemed to glow.
  • The power measured at the end of the output fiber (fiber after the waveguide) was 0.6 mW. Not entirely sure what the contribution of loss was in the decrease from 2.74 mW through the waveguide.

• The laser is still ON although the shutters to the green and IR paths are closed. Safety glasses required before opening shutters.

Questions about the setup

1. The spec sheet on the wiki mentioned PM980 and PM480 input and output fibers, respectively for the waveguide operating as an SHG, what were being used instead were P3-1064PM-FC2 and P3-488PM-FC2. Is this a significant source of loss that can be easily remedied?
2. Yehonathan mentioned that stimulated Brilluoin scattering occurs in all fibers above a threshold. What is the threshold for the the ones used in the setup? We probably want to operate below this threshold.

Our next step would be to measure the LO shot noise.

{Shruti, Yehonathan}

We made some a list of some random questions and plans for the future. We then went down and found answers to some of those:

1. Why is there no Faraday isolator in the 1064nm beam path? (edit: turns out there is, but inside the laser, see pictures in this elog).

2. Do the fibers joined by butt-coupling have similar mode field diameter? If not it can explain many loss issues.

a. In the green path we find that according to the SPDC datasheet the 532nm fiber (coastalcon PM480) is 4um while the input thorlabs fiber (P3-488PM-FC2) coupled to it has an MFD of 3.3um. This mismatch gives maximum coupling efficiency of 96%. Ok not a big issue.

b. At the 1064nm output the SPDC fiber is PM980 with MFD of 6.6um while the BS fiber is 6.2um which is good.

3. What is the green fiber laser damage threshold? According to Thorlabs it is theoretically 1MW/cm^2 practically 250kW/cm^2 for glass air interface. With 3.3um MFD the theoretical damage threshold is ~ 80mW and practically  ~ 20mW. It doesn't sounds like a lot. More so given that we could only get 50% coupling efficiency. How much is needed for observable squeezing? There is the possibility to splice the fiber to an end cap to increase power handling capabilities if needed.

4. Is stimulated Brillouin back scattering relevant in our experiment? According to this rp photonics article not really.

5. How much green light is left after the dichroic mirrors? Is it below the shot noise level? Should check later.

In addition, we found that the green fiber input and the 1064nm fiber output from the SPDC were very dirty! We cleaned them with a Thorlabs universal fiber connector cleaner.

[Yehonathan, Shruti]

1. Doubling cavity and green beam

Since we had left the lasers ON with the shutters closed we wanted to see if the powers measured after opening the shutter would be similar to what it was when we left. We realized that opening and closing the green shutter destabilizes the doubling cavity (the FI is after the shutter and the shutter does not seem to be a good dump), which in turn changes the SHG crystal temperature (possibly because of the power fluctuation within the crystal). Re-opening the shutter requires some tuning of the temperature and offset to recover similar output power. Finally, after some tuning, we were able to see 156 mW of green light.

2. Attempt at measuring LO shot noise

• We want to measure the BHD output A-B channel, which we expect to be dominated by shot noise since all the classical noise would be canceled when optimally balanced, but found that one of the PDs was inverted so that the sum of the two channels would be what we needed to measure. Since the SR 560 has only an A-B option, we used a second SR 560 to invert the B channel before subtracting
• Operating at an LO power of ~4 mW did not give us sufficient clearance from the dark noise in each channel with the SR 560s, which was around the max power I believe we're supposed to use for the BHD LO
• Somehow measuring the output directly, without the SR 560, gave us some clearance over the dark noise at ~1 MHz and higher (possibly because 1 MHz is the SR560's BW) so we decided to measure the time series of both channels together and do the optimized subtraction and FFT offline

3. Subtracted noise signal spectra [Attachment 1]

• The plots show the noise spectra of the channels measured individually. The 'gain adjusted' means that A was multiplied by 1.05 and B by 0.95 in order to get the two plots to more or less line up
• We used the Moku to measure the timeseries at a sampling rate of 10.2 MS/s for a period of 1.2 ms with AC coupling and 50 Ohm impedance. Elog 2324 suggests the designed measurement was for a 50 Ohm load so we should be impedance matched but I'm yet to convince myself
• Our estimate for the shot noise was 77 nV/rtHz for 4 mW of power and 87% QE, using a TI gain of 2kOhm (the black dashed line in the plot). If we were impedance matched the yellow trace must be higher than this estimate
• In our next measurements, we will also record the dark noise, carefully measure the power. There is obviously sonething wrong with the plot

30 Mar 22 edit: script here, data here

[awade]

Good to see this experiment being revived.

1. The design of this laser had a number of flaws and one of them is this sensitivity to backreflections at 532 nm. I mostly just disabled the doubler's lock and closed the shutter for good measure, but probably best not to leave flickering around in an unstable state when you're away.

2. I built in the inversion in the second channel to give myself the option to electronically subtract: something that didn't end up being very practical compared to just digitally recording channels and subtracting in post.

• I'm surprized the SR560 don't given you clearance there.  Nominally these units are 4 nV/rtHz if you operate in low noise mode AND gain=100 (see p21 of the manual), below this gain gives you 10-60 nV/rtHz noise. When I built the PD circuits I did verify that I was getting the clearanceI expected.  At 1 mW on each photodiode one would expect order =sqrt(2*h*c/lambda0*Responisvity*Power)*Gain ~ sqrt(2*6.626e-34*3e8/1064e-9*1e-3*0.7)*2e3 = 32 nV/rtHz.  
• A few things to verify:
  • check the DC voltage (just with a multimeter) to see true power picked up by diodes, this should be 1.4 V for about 1 mW of 1064 nm;
  • Make sure you're AC coupled into SR560, there is no way you operate at gain 100 or above and also not saturate for 1 mW (~1.4 V amp output DC) of light
  • at 2kΩ gain you should expect the noise floor to be of order =sqrt(4*kB*T*G)=sqrt(4*1.38e-23*300*2e3)~5.8 nV.  Only just clear of the SR560 spec, and about equal to typical actual performance.  To see this level you might want to pre-amplify with a Femto amplifier, the 40m Busby box or the ganged amplifier box I made for the CTN lab (its black with gold writing, Anchal knows the one). A dark measrument like this may have a little offset that you can either null or just AC couple with a minicircuits DC block;
  • Take a terminated (50 Ω) measurment of your ADCs when you collect your PD 'dark' data. Even better also collect terminated SR560 data. And put these on the plot.  Moku:Labs have ~30 nV/rtHz @100 kHz and above.  Just be sure you're measure photodetectors and not Pre-amp or ADC noise. Moku nominal input refered noise is 13.9 nV/rtHz * sqrt(1+220kHz/f).
  • If you can't get any quick progress, try all the above with minicircuits lower noise amplifiers.  They have plenty of bandwidth and go to higher frequencies.
• Just measureing the output of the PD directly, with no subtraction or amplification, I'd say you are looking at laser technical noise at about 1 MHz: this is what the subtraction is for, to null the LO noise effects to only listen to signal port.  Maybe somethings burried in the subtracted signal offline, but you need some simultaneous termianated measurements back long the signal chain to put some bonds on what is the limiting noise here.

3. Subtracted noise spectra

• These gain numbers sound right to me.
• The AD829 is designed to drive this combined 150 Ω load, I would stick to 50 Ω terminations for now. On the topic of mokus: again verify its input referred noise and pre-amplify accordingly.  Also there is a choice of "Normal" or "Precision" acquisition mode, I think the right choice is precision (this should have the right filters to kill aliasing from the downsampling)
• ~70 nV/rtHz shot noise sound about right to me.  Not clear why a subtracted signal doesn't seem to reach this.  Once again, measure the actual DC voltage output from the TIA to get the true absorbed photons and use that to calibrate your estimate.  

We should chat some time on zoom about more details (rana can forward my details).  Hope this enought to go on for at least the homodyne part of the experiment.  

{Shruti, Yehonathan}

Yesterday, we measured a bunch of noises.

We wanted to have as reference the Moku noise, the PDs noise, and measure the shot noise of the LO again.

Attachment 1 shows the Moku noise measured by just taking data with no signal coming in. We tried both the spectrum analyzer (SA) and the oscilloscope tools, with and without averaging, and the difference between the channels.

For some reason, the SA has a worse noise figure than the oscilloscope and the difference channel doesn't give us any special common-mode rejection. Also more averaging doesn't help much because we are already taking 1.2ms of data which is way longer than 1/RBW=0.2ms we are taking here.

From now on we use the oscilloscope as the spectrum analyzer and to its noise we refer as the Moku noise floor.

Moving on, we try to measure the PD dark noise. Given that the PD dark noise floor is ~ 6nV we don't expect to see it with the Moku without amplification. Attachment 2 shows that indeed we couldn't resolve the PD dark noise.

We then opened the LO shutter. We measured with a power meter 1mW and 1.15mW coming impinging on the PDs. The voltage readings after the preamp were 1.66V for the white fiber, and 1.93 V for the red fiber. These values suggest responsivities of 0.830 and 0.834 respectively.

The PDs were measured using the Moku scope and subtracted digitally with some small gain adjustment (0.93*ch1-1.07*ch2) between the channels. The result is shown in attachment 3 together with the expected shot noise level.

1. There is not enough clearance for detecting squeezing.

2. Expected shot noise level is still too high. Does the 2kohm preamp gain go all the way above 1MHz??

[Yehonathan, Shruti]

Yesterday we went back to fiddling with the green path. Soon after opening the green shutter and then switching the doubling cavity to 'AUTO' we were able to see 150 mW of green light. We were able to replicate this a couple of times yesterday.

Since we had earlier removed the green fiber from the fiber launch to clean its tip, the coupling into the fiber turned out to be quite poor. As can be seen in Attachment 1, Yehonathan pointed out that a lot of green light was being lost to the cladding due to poor coupling. He then played around with the alignment and finally was able to see 65% coupling efficiency. This process seemed to involve a great amount of trial and error through several local power minima.

Attachment 2 shows that the coupling between the two fibers at the 532 nm input of the waveguide is quite poor (there is visible light being lost in the cladding). Furthermore, this light intensity decreases as we get closer to the waveguide meaning this light is being dissipated in the fiber. Even at the 1064 nm output where we expect to see squeezing there is some remnant green light.

We wanted to test if the green leakage reaching the PDs were causing additional noise. For this we just looked at the spectrum analyzer on the Moku (after amplifying 100x with the SR 560) and saw no difference in the noise spectrum with and without the green shutter being open. Although, we're not convinced with this measurement since we were not able to find good quality SMA cables for the entire path. Moving around the BNCs seemed to change the noise. Also, near the end, we noticed some coupling between the two channels on the Moku while measuring the noise that seemed to cause additional noise in one of the channels. We did not have sufficient time yesterday to probe this further.

{Yehonathan, Shruti}

1. Grabbed 30Hz-3GHz HP spectrum analyzer from the Cryolab. Installed it in the WOPO lab under the optical table. We figured out how to do a zero-span measurement around 10MHz. The SA has only one input so we try to combine the signals with an RF splitter. We test this capability by sourcing the RF splitter with 10MHz 4Vpp sine waves from a function generator and measuring the output with a scope. We measure with the scope 1.44Vpp for each channel. The combined channel was 2.73Vpp. We then realized that we still don't have a way to adjust the gains electronically, so we moved on to trying the RF amplifiers (ZFL500 LN).

We assemble two amps on the two sides of a metal heatsink. We solder their DC inputs such that they are powered with the same wire (Attachment 1). We attach the heatsink to the optical table with an L bracket (Attachment 2).

We powered the amps using a 15V DC power supply and tested them by feeding them with 10MHz 10mVpp sine waves from a function generator. We observe on a scope an amplification by a factor of ~ 22. Which makes a power amplification of ~ 26db consistent with the amplifiers' datasheet.

We couldn't find highpass filters with a cutoff around 1MHz, so we resumed using the DC blocks, we test them by feeding white noise into them with a function generator and observing the resulting spectrum. First, we try the DC blocks with a 50 Ohm resistor in parallel. That happened to just cut the power by half. We ditch the resistor and get almost unity transmission above 20kHz.

Moving on to observing LO shot noise, we open the laser shutter. We find there is only 0.7mW coming out of each port of the fiber BHD BS. We measure the power going into the BS to be 4mW. This means the coupling between the LO fiber and the BS fiber is bad. We inspect the fibers and find a big piece of junk on the BS fiber core. We also find a small particle on the LO fiber side. We cleaned both fibers and after butt coupling them we measure 1.6mW at each port. We raise this power to 2mW per port.

We connect the outputs of the PDs to the amps through the DC blocks. The outputs of the amps were connected to the Moku's inputs. The PDs were responding very badly and their noise was also bad. We bypass the amps to debug what is going on. We connect the PDs to a scope. We see they have 300mV (attachment 3) dark noise which is super bad and that they hardly respond to the light impinging on them (attachment 4). We shall investigate tomorrow.

[Yehonathan, Shruti]

First we turned on the relevant instruments for this experiment after the power shutdown:

- Main laser drivers and doubling cavity controller. We set the current to 2 A as we had it before.

- The waveguide TEC. We tried setting it to 60.99 C (for maximum efficiency) but the temperature ramps up much faster and over shoots the setpoint. So we had to do what we did earlier which was to adiabatically change the setpoint from room temperature and finally set it to something like 63 C so the actual measured temperature stabilizes at ~60.9 C. How do we change the PID parameters on this controller? The settings don't seem to allow for it.

- PD power supply, oscilloscopes, function generator, SR 560s lying nearby

Then we tried to probe further what was going on with the PDs (TL;DR not much made sense or was reproducible): 

• Initially we were sending in 12 V from Ch1 of the Tenma power supply and 15 V to the mini-circuits RF amplifiers from Ch2 of the same Tenma. When we tried to observe the DC voltage levels (before amplification) and the noise (after amplification) but they did not make sense to us as described in the previous elog.
• Then we disconnected the RF amplifiers and switched the power supply for the PD transimpedance amps to Ch2 of the Tenma. Briefly we were able to see 2 V DC at each output (as expected for the ~1.3 mW of light incident  on each PD) when measured with the multimeter. On the scope we were seeing that one of the PDs was dominated by 60 Hz fluctuations with a much lower
• We tried to switch it back to Ch1 but even the DC levels seen by the multimeter were bogus (~1 V for one of the channels and 2 V for the other)
• When we switched the power supply to a HP one taken from CTN, the DC levels on both channels seemed ~1V without any light and the noise somehow seemed lower with some incident light

Possible next steps

• Even though the AD 829 data sheet says it can be operated at up to 15 V, it is possible that we damaged both of them somehow when cycling the power. Elog 2324 also says that it was designed to be powered at 5 V. We could replace both op-amps and measure teh transimpedance TFs before and after the change
• Use different PDs. Maybe temporarily try with just the ThorLabs PDA20CS or similar with ~20 mW of LO power and measure the shot noise (possibly also squeezing).

{Shruti, Yehonathan}

We realized that the PD amp circuit only requires a 5V DC supply so we try that. One of the PD had the right response, although only after cycling the input impedance from 50ohm to 1Mohm which is weird. The other one (which produces the negative signal) was complete bonkers.

We remove the home-built PDs and put 2 Thorlabs PDs (forgot the model) with a bad dark current but a decent response and high saturation current. With these PDs we are limited by the PD noise to about 1.25db od squeezing when 30mW LO is detected on each PD without using electronic amplifiers. Attachment 1 shows the different noise spectra we measured.

We maximize the coupling efficiency before boosting the LO power. For some reason, the coupling between the LO fiber and fiber BS deteriorated but there was no apparent dirt on them upon inspection. We crank up the power and measure PD outputs using the Moku oscilloscope. The PD signals were subtracted digitally, but now we were not able to get the shot noise even after fine-tuning the gains. What went wrong? maybe it's because the PDs have separate power supplies?

Details later...

Some analysis in this notebook

{Shruti, Yehonathan}

We went to the e-shop to investigate the PD circuits. Completely confused about the behavior of the PDs we decide to gain some sanity by testing a sample AD829 on a breadboard with resistors and capacitors similar to those in the design of the PD circuits shown here. The PD is replaced by a voltage source and a 2kOhm resistor such that 0db gain is expected. We first measure the TF of the opamp with the Moku just with the resistors (attachment 1) then with the compensation capacitors.

We tried powering the opamp with shorting V- to ground like we did in the WOPO lab (for some reason this was how it was connected) and got garbage results (attachment 3).

We then turned to retesting the PD circuits with a proper powering scheme. However, connecting +/-5V and ground from a power supply resulted in the output of the PD circuit being ~ -2V even when the PD is taken out which might suggest that the opamps have really gone bad.

[Shruti, Yehonathan]

We took newport 1811 PDs, one from CTN lab (suspicious) and one from (I forgot) for their high gain and low dark noise.

The detector diameter is small 0.3mm, but our focusing is sufficient:

The mode field diameter of the PM980 fiber is ~ 6.6um. The beam is collimated by a Thorlabs F240APC-1064 with f = 8.07mm and focused with an f = 30mm lens. It means that the diameter at the focus should be roughly 6.6um*30/8.07 = 0.024mm which is well within the PD active area.

We place the PDs at the focal point of the lens at the BHD readout. The impinging optical power was set to be ~ 0.6mW at each port. In one of the PDs, we measure the DC response with a scope to be ~ 5.5 V/0.6 mW ~ 9e3 V/W. According to the specs, the DC monitor as a response of 1e4 V/A while the responsivity of the PD itself is ~ 0.8 A/W at 1064nm so the overall responsivity is ~ 8e3 V/W.

However, the second PD's DC response was bonkers: we measured it to be ten times less. The AC response might still be OK since it is a different port but we haven't measured it yet.

{Yehonathan, Paco}

We comfirmed that the DC ouput of one of the 1811s is bad. We set out to measure the AC response of the PDs.

For this, we decided to use the current modulation on the Diabolo laser which is rated to have 0.1 A/V and a bandwidth of 5kHz. We calibrated the current to optical power by swiping the current and measuring the power at the homodyne PDs using a power meter. The laser power before the 1064nm PZT mirror was measured to be 5mW.

Attachment shows the measurement and a linear fit with slope=0.97 mW/A.

We drove the current modulator using a sine wave from a function generator with 1kHz 0.5Vpp. When we looked on the PD AC signal port in the Oscilloscope we saw 2 Vpp 12Mhz signal. We passed the signal with a low pass filter but again we saw mostly noise.

We took the PDs to the 40m PD test stand but we accidently fried Jenne's laser.

Next, we should just use the Moku network analyzer instead of the scope to measure the response in the QIL using the Diabolo current modulator.

I recall at one point we had one of these NF1811 with a broken power suply pin.  It was from a limited production run with the smaller micro 3-pin power connectors. Maybe check yours is not that one.  

Long story short it still responded with only the positive rail but DC will gave a bad photovoltaic mode response and the AC had a large unstable oscillation that was only viewable on a high speed scope (if I recall right higher than the 125 MHz nominal bandwidth).  I would check the power-in pins aren't bent/broken and also check the AC out on a higher speed scope (i.e. >=500 MHz).

{Shruti, Yehonathan}

While setting up a new green coupling set up a spacer fell on the bare waveguide input fiber and broke it.

The broken piece is not very large and we still have a lot of fiber (attachment 1) to work with.

Attachment 2 shows the mode of the broken fiber illuminated with SHG from the waveguide.

Next Steps

We need to find a fiber stripper and a cleaver. This is a common tool in fiber optics labs and I'm sure we can find a place on campus to borrow from. If we don't find it we can buy a fiber stripper (like this) and a cleaver (like this) from Thorlabs.

Once we have a cleaved fiber end we are back in business.

In the longterm, we will have to make a new connector for the fiber using one of these kits. Something I have some experience with.

This happened mysteriously and had absolutely nothing to with me. The fact that I was the last person to open the filing cabinet before this happened is circumstantial and beside the point.

Will get the lockshop onto this in the next couple of days. In the meantime, just try and exercise your clairvoyance.

Fixed! The key is now hanging in the second bay from the right above the main bench.

I tried to make a change to the front-end in Simulink and compile it on fb0 - which is supposedly now our front-end machine. For some reason it won't build the .rtl file that is the front-end itself. When I tried to revert to the backed-up model I had the same problem. It doesn't look like anyone has tried to rebuild the front-end since late September and there have been some changes to the network since then.

I'm going to track down Alex and sort this out.

Aidan and I began circling the holes in the ventilation system with Sharpies this afternoon, only to find that the situation was even worse than we thought before (which was already bad). Below is a picture of one particularly bad section (compare with Frank's previous post). We can continue to highlight each one, but by the time we are done the entire lab will look like this--maybe I should have chosen 'fugly'. The worst leaks are not the small ones that can be seen on the sides and bottom of the ducts, but rather the HUGE gaping holes on the top sections of nearly every joint in the system. I am fairly sure that even if we were to highlight each and every hole, we will not get PMA to fix each one. Rana suggests that we at least get them in to fix the largest ones for the time being.

I was attempting to replicate the pwelch command using the numerical recipes formula for the Welch's Periodogram PSD estimate. Koji helped me figure out that the calculation was missing a factor of bandwidth, and even though it said explicitly it was a PSD, it was a power spectrum. He also helped me figure out some stupid errors I was making in indexing.

I have recreated the algorithm now, and am going to use this, along with TFestimate to do frequency domain subtraction. I think this could still be a good thing even with the MZWino code, but it will be good to at least compare.

I restarted the ELOG on NODUS just now. Our attempt to set up error logging worked - it turns out ELOG was choking on the .ps file attachment.

So for the near future: NO MORE .PS files! Use PDF - move into the 20th century at least.

matlab can directly make either PNG or PDF files for you, you can also use various other conversion tools on the web.

Of course, it would be nice if nodus could handle .ps, but its a Solaris machine and I don't feel like debugging this. Eventually, we'll give him away and make the new nodus a Linux box, but that day is not today.

To restart the elog:  http://lhocds.ligo-wa.caltech.edu:8000/40m/How_To

 - apparently PDFs work okay. Frank is reporting continual crashes from last night when uploading a graph of the particle counter in different formats from Firefox in Windows.