Sorry, you're right that must be the wrong GPS time I've got there. It was stepped back down on Friday around 7-8pm.
Also, the output impedance of the SR560 is going to be 50ohm (not 100 as it was for the buffer amps.) Maybe you can either add the 294k onto the values in your graph, or label it as being delta_T.
Stepped lab temperature up on both controllers by 5K at 994195123 GPS.
Stepped lab temperature down at 994195351
These are only a few minutes apart, right? It looks from this data (outside the foam box in blue, inside in red) like nothing really happens at 994195123 but the temperature goes down at ~994197123...
Added x50 gain using two SR560s. I put 2.94v into the B input and set the amps to A-B such that the offset voltage is removed. We can't add much more gain than this since the SR560 can only do +/-5V and we need to have room for the temperature swing when we step the lab temp.
Stepped lab temperature up at 994454835
Stepped lab temperature down at 994462690
??....this looks like its all ADC noise.
How about a plot showing the spectrum of the ADC noise and the sensor noise and the measured temperature noise all on one plot with the y-axis in units of temperature (i.e. not cubits/sqrt(furlong))
This morning we looked at the response of the temperature sensors inside and outside the foam box to yesterday's ~5K temperature decrease in the room. There was a weird square (ish) noise signal in the first hour or so of data: the temperature was increasing and decreasing by the equivalent of 1 bit in the DAQ every ~10 minutes. This could have come from a number of places--Frank suggested that it might be a fan in the DAQ turning on and off, either heating/cooling the cables or coupling into the electrical signal. It could have been the air conditioner turning on and off, but it's weird that such fast fluctuations would affect the temperature inside the box as well as outside. Or, most likely (especially because the ups and downs were 1 bit) it could be random DAQ noise. This afternoon we investigated possible sources of the problem by putting a BNC shorting plug on the input where the temperature sensor would connect and measuring just DAQ noise. The attached PDF shows the DAQ noise (in green) and the temperature outside the box that we measured this afternoon. For reference it also shows the temperature inside and outside the box at around 4am.
I took a transfer function of the first assembled PDH2 box (given S/N 7371). I was able to do it open-loop using the random-noise source of the agilent, high-pass filtered through an SR560 (the same way as in this post). Here it is along with the LISO estimate:
I believe the low-frequency discrepancy might be explained by occasional saturation of the output. In any case, this is just a trial filter; the one we actually implement is likely to have a very different shape, especially since we will soon replace the cavity mirrors and move the pole down.
Here is the input-referred noise level and the LISO estimate:
The high-frequency weirdness is just due to the noise floor of the analyzer. To be convinced, see the output noise spectrum below. The low-frequency part is more troubling, but it is the same sort of effect we saw with the temporary breadboard version (see link in first paragraph). In any case, it should wind up being below our goal noise level, even before the optical gain enhancement.
I think the next step is to set up the servo in a configuration best compatible with the current optical setup, and actually replace the uPDH boxes with PDH2s. Then, we can do in-situ testing.
Just stepped the temperature down by ~5K, at 5:33pm or 994034012 GPS time.
Also, we changed the sample rate on the 4 test input channels on the DAQ down to 256 Hz, and gave each one a 4th order low pass filter at 1Hz.
We made a new file in order to record epics data for the ATF. It is
and we added it to the master file for the framebuilder
At the moment it just has these 4 channels in it, but we rebooted the framebuilder and can now get past data from the slow channels.
We have just stepped the temperature in the lab up by ~5K, at 12:20pm (GPS=994015129)
I finished stuffing and testing our two PDH2 servos. I went through taming the AD829s one stage at a time, installing compensation caps as per the datasheet. There are no more oscillations and the noise levels throughout are as simulated by LISO.
The remote sequential boost control is also working properly, as seen below.
I have assembled one full NIM module and it looks pretty good. Some pictures:
Here is a demonstration of the boost switching. In the first picture, the control voltage is 2V, putting the servo in state 2 with two boosts engaged. In the second, the voltage is 3V, so there are 3 engaged. You can see the indicators reflecting the switch (because of a last minute change to the circuit, it ended up that the lights are OFF when the boosts are engaged---I switched them to red LEDs so that it seems more intuitive, as all the lights will be off when things are functioning normally).
Here is a shot of the module in the crate, with the other board resting in it to prove it's also done :-)
The next steps are:
HAPPY 4th OF JULY!
Here are the temperature measurements from last night. Red/test1 is taped directly to the optics table. Blue/test2 is in the box on the table. Green/test3 is next to red, but in the air. Brown/test4 is inside the mirror cavity of the gyro. As you can see, the sensor in the air sees relatively large fluctuations over long time scales/at low frequencies and the one in the mirror cavity sees some fast fluctuations (which might be caused by a glitch other than actual temperature change?). The one on the bench is very stable, presumably because it is touching a large volume of metal which has some heat capacitance. The one in the box is also stable, which might mean that the box is helping to insulate from temperature fluctuations in the outside air.
1. The time series data only shows up in short (140 second) chunks. There is about 10 times as much data as is shown on the plot, but in order to show all of it at once we'd have to mess around in MATLAB for a while. Alastair plans to show me how do this next week.
2. The low input impedance of the ADC is still affecting our data. It's about 2k and the output impedance of our amplifiers is 100 ohms. This means that it's affecting our data by a few percent, which on the whole Kelvin scale comes out to something on the order of 10 degrees. Since we're measuring temperatures of around 20 degrees C, this becomes a fairly significant problem. For now I just corrected for it with a multiplying factor in the calibration based on the input impedance that we measured last week, 1977 ohms. The next step will be to take a more accurate measurement of the input impedance at each channel. Still, even after we have the accurate measurements we will probably want to fiddle with the electronics to fix the problem more thoroughly. Does anyone have a good intuition for how to do this?
I'm trying to design a generic circuit board that will go in a NIM crate in the gyro lab and allow for a flexible configuration of a number of different filters. I just finished a first draft of a schematic and it would be fantastic if y'all could look it over. Please let me know if you notice any mistakes, have questions, or see ways that it could be improved!
This design has a power source, voltage regulators, 3 twin T notch filters, 5 generic filters and a differential amplifier (and I'm pretty sure all of that will fit on the PCB). The generic filters are sort of complicated looking, so I also uploaded a drawing of a few of their possible configurations. Basically, they just have a lot of 0k resistors and 0F capacitors so that on each footprint you can choose to put in nothing, a 0k resistor, or the resistor/capacitor that you need depending on the design. Most of the components/footprints and some of the filter designs are taken from schematics by Zach, Alastair and Rich Abbott.
There are 4 sensors (photo 1 below) measuring temperature around the laser gyroscope right now. They are hooked up to a little black box (photo 2 below) that Alastair and I finished building today. We are taking data using channels 17, 18, 19 and 20 in the ADC. Channels 17 and 18 were already set up to record data but we had to add 19 and 20, so we did restart daqd and DAQ reload when setting those up around 5:30 pm today.
Because of the fairly low input impedance of the ADC we build a buffer amplifier for our temperature sensors. At the moment it just has four channels (this was the size of the board we used) but we can expand it later if we need. The output of the temperature sensors is around .29V at room temp so the buffer amplifier also gives some gain ( x 9.82). Inside it is just four non-inverting amps using OP27 along with 15v voltage regulators that run off the 24v of the NIM crate. The voltage regulators are bypassed with caps (0.015uF, 1uF and 100uF) and diodes. Their output impedance is set to 100ohms. It's all packaged inside a NIM crate and we added an LED to make it look pretty.
We tested the voltage regulators and they powered up fine, then we tested each amp with a signal generator and scope. They all function just as we expect.
We measured the input inpedance of the ADC in the ATF, using a battery and a range of known external resistors. The value comes out to be 1970ohms. This is worth remembering since we don't want to assume it has high input impedance.
I finished purchasing all the components for the PDH2 today, including the RF stuff from MiniCircuits and CoilCraft, SMA/SMP adapters and miscellaneous switches/LEDs/pots/etc from DigiKey, Allied and Mouser. I also ordered 6 NIM front panels from FrontPanel Express to go with the 6 boards that were ordered (in production now) form Sunstone last week. Lead times will vary I suppose, but everything I ordered was in stock.
EDIT: I changed the offset trimpot to a horizontal-mount one accessible through the front panel. I also verified clearance with the larger gain adjust pot's external dial.
I figured out where the dimension confusion was coming from and fixed the errors in the front panel design. This allowed me to pick appropriate places for the I/O headers on the board (so that they would have clearance from the BNCs on the panel). I also investigated some other clearance concerns, such as with the potentiometer knobs on the outside of the panel, as well as between the individual BNC connectors on the panel. Finally, I rearranged some stuff to make sure there was enough room for the heads of the screws that will be used to mount the board via standoffs to the NIM module.
Here are the current PCB layout, schematic (unchanged), and NIM front panel. I think we are ready to start ordering stuff.
I was finally able to reconcile the PCB design with Frank's latest comments. The basic changes:
In addition, Rana suggested putting pads for capacitors at the low point of the RF coupling transformers because this helps with pickup somehow(?). I also added pads for resistors at the high points, so that we could just not stuff the transformers and short the low and high points across to directly couple should we decide we want to.
Attached are the current schematic and PCB layout. I realized that something was off with the board dimensions, so the audio I/O headers are not in their final locations yet. I am trying to place them so that they aren't in the way of the BNC mounts on the panel.
I have run the design rule check on the new PCB and there are no violations. I am going to do one last run of checking dimensions (hole placement, BNC mount clearance, etc.) and then generate files. Once I have them, provided there are no more comments, I will take them to Steve or someone to purchase.
I've put two Coilcraft 1:1 transformers (PWB-1010L) in the schematic, one between the LO input and the mixer and one between two SMP connectors (in from the RFPD and out to the mixer RF input). You can see the schematic if you click on the PCB below to see the full file.
I guess that we should transformer couple the LO and RF inputs to the box in any case. There's no downside to this. Its better to do it now, before stuffing the board rather than try to hang an external transformer off of it later. Make sure to get a 1:1 with a high enough power rating.
We are going to power the servos and the PDs from the NIM crate. Rich thinks that there shouldn't be a problem since there is very little area within the loop (because the power supply cables and RF signals travel next to each other from the table to the rack). If we have a problem we will transformer couple the inputs to the mixers from the PDs.
the originial power supply already has +/-15V, that's why there are 12V regulators inside. If you wanna change it to 15V regulators in the box you have to replace the power supply too. If you want to use a centralized power source then you need differential or transformer coupled inputs/outputs to remove ground loops
Huh? I'm not sure I understand the significance of going from +/-12V to +/-15V. Is one special?
I made a NIM front panel for the PDH2. The only thing that's left to work out are the catchy screws that go on the sides to hold it to the crate. The price is $59.66 for one or $53.69/ea for 5-9.
you've changed the supply from =/-12V to +/-15V. How do you wanna supply it? Using your existing +/-24V supply will create lot's of ground loops as there are no differential inputs or transformer isolated inputs/outputs for the RF stuff.
I have made the recommended changes in our email thread and gone through an extra iteration of changes with Koji this afternoon. The current schematic is below.
I feel good about the board at this point, so I'm starting on the PCB. Please let me know ASAP if there is a major problem, and send me any minor concerns/comments whenever.
I finished with the layout, including the RF stuff (plus a spot for the mixer) and the testpoints. I have used headers for the audio I/O which we can use to connect the board to the BNC mounts on the panel. I will get with Alastair to do the final design check, since he's done it before.
In the meantime, I will put together a panel model with FrontPanel.
All but finished with the all-new PCB layout for the PDH2. Still need to wire up the RF stuff and work out how we want the connections (these will be a mix of SMP and twisted pairs from the board to the front panel of the NIM module to avoid stress and over-specification). The open central region of the board is for the minicircuits mixer.
I have made most of the suggested changes to the PDH2 schematic. The original comment thread can be found here.
What I changed:
What I have not (or not yet) changed:
Here is the schematic. As soon as I have fielded some more comments I can move on to redoing the PCB.
[Koji, Alastair, Zach]
In taking measurements to make a current noise budget for the gyro, we noticed that the primary error signal spectrum had a suspiciously similar shape to the gyro noise at low frequencies. We realized that we could now be sitting on the "spillover noise" that was less than an order of magnitude below our sensitivity before the upgrade. As a reminder, this is common-mode noise that is not fully cancelled by the primary loop (since it doesn't have infinite gain) and thus enters as residual noise into the secondary loop, looking like a differential (i.e. gyro) signal. It is calculated by simply taking the primary error signal and converting into angular velocity noise by dividing by the optical gain and multiplying by the standard Hz -> rad/s factor.
We decided that we should try adding low-frequency gain using two SR560s (one as a one-pole LPF and the other flat gain, summed together). We increased the gain by a factor of ~100 in the gyro band, but only observed an improvement of ~10 in both the gyro noise and error signal spectra. Below are plots of each. This is strange, but it does suggest that the limiting gyro noise source comes from somewhere in the primary loop, so that narrows some things down. Figuring that our SR560 rig was too hacky, we tried reinstalling my second boost switch by changing R30 to a 110k resistor and then putting a switchable leaded 1k resistor in parallel. The transfer function was what it should be, but the noise improvement we saw was not as good as the SR560 approach.
After taking loop transfer functions to determine the optical gain, I calibrated the primary error spectrum into equivalent spillover noise, both in the standard configuration and with the SR560s in for extra gain. They are plotted below:
This points out something interesting: even though the extra LF gain has roughly the same effect on both the gyro noise and primary error spectra, the spillover noise does not appear to be high enough to explain the gyro noise at low frequencies. The agreement at a few hundred Hz convinces me that the calibration is about right. The similarities in the LF spectra are strong enough that I am fairly certain the gyro noise comes from the same original disturbance as seen in the error spectrum, only it can't couple in via the spillover noise pathway. There must be some other coupling, and perhaps the fact that the extra LF gain has the observed effect on the gyro signal is a good place to start thinking about what it is.
We have been trying to figure out why we see (more) excess low frequency noise in the current configuration. The ~100x optical gain imbalance could have been an unrelated issue, but it seemed like a reasonable place to start today.
We began by bypassing the PDs altogether and just injecting part of the LO signal directly into the RF input of the mixers (where the RFPD signal comes in). This should create a DC signal at the low-passed IF output, which can be maximized by tuning the LO phases with the dials on the boxes. Using equal input RF levels, the maximized outputs of the respective mixers were different by a factor of ~100 (220 mV to 2 mV). We double-checked that the LO levels were equal, and since they were we deduced that something was wrong with the primary box's demod electronics. Checking each of the parts individually, we found that the ZX05 mixer was busted. We replaced it and the optical gain is now 100x higher. This explains the discrepancy we have been noticing from the theoretically achievable level for months.
I reduced the gain of the servo to accommodate the optical gain, but we still needed >10dB of attenuation between the PD and the mixer to keep the loop from oscillating. This needs investigation.
When we plugged everything back in and took a spectrum, we found that the low-frequency noise was back to the level (a bit lower, in fact) where it was before we reduced the optical power from 30 mW to 3 mW. It is a bit worse at higher frequencies, but this is the best we have ever done in the current configuration. Plot:
It is a bit disheartening that there isn't a real improvement, but we must consider this:
Yesterday we did some loop characterization to see if the increased LF noise we are seeing could be from electronics noise. First we measured the OLTFs:
We also verified that the servo transfer functions are what they are supposed to be. They were. Here are the LISO TFs:
Using the above two pieces of information, along with:
We can compute the optical gains ( [V/Hz] ) to be:
There is an equal power of about 3 mW in each direction, and the REFL PDs are identical, so it is very confusing that the optical gains are different by more than a factor of 100!! This needs to be investigated immediately.
Taking the above as true, we can still calculate the current gyro noise contributions by taking the output noise level (with the input terminated), referring to the input, dividing by the optical gain (and 3 dB attenuation for CCW) and multiplying by the standard Hz->rad/s factor. The result of this is below:
Comparing to the red curve in the plot here, we see that the electronics noise is not high enough to explain what we see in this configuration. We still need to check the dark PD noise and demod noise in this configuration, though.
[Alastair, Koji, Zach]
We had some progress today, which can be summarized in a few parts:
1. Modulation depth
Either I measured this wrong in the previous entry or this magically got better. Koji and I measured it together and we saw sidebands at the level of ~60 mV to the carrier's ~4 V, giving
Γ = 2 sqrt(60 mV / 4 V) = 0.24 rads This is GOOD.
2. Spurious AM
Continuing with what Alastair started this morning, we tried to reduce the unwanted 19 MHz junk in the PD signal. Alastair pointed out that there was a reasonable amount of noise even with the shutter closed, suggesting some sort of electrical pickup. We tried making some changes to the distribution of the LO signal to the EOM and found that separating it in space from the REFL PD signals returning to the rack made it much better. We also noted that using an SMA cable (instead of BNC with an adapter) for the final stretch over the table to the EOM reduced the pickup further. Koji suggested that I steal a long SMA cable from the 40m so that I could use it for the whole stretch.
3. LF noise
With these improvements the gyro was easily lockable again, and we measured the noise, which we found to be of the same (terrible) level as the other day. We decided to try installing the Cougars into the REFL PDs to add 10 dB of optical gain. This was in part because we had reduced the optical power from >30 mW to 3 mW and figured we could use the gain up front, and in part because we knew that reducing the optical power was one of the only things we had done between now and when our LF noise was lower. I also changed R23 of PDH box #1437 again, this time from 220 to 1k to reduce its gain by ~5 (enough to counteract the added 10 dB of PD gain plus a little more because we were already having to attenuate the PD signal to keep the loop stable).
The result was that the LF f-1/2 noise was shifted down by a little over a factor of 10 (see below). This suggests that the current limiting noise source is probably in the servo electronics and is made less important by shifting the gain to the PDs. I imagine that increasing the optical power while reducing servo gain further will cause it to go down even more (though we shouldn't do this since we want to keep out power at ~3 mW). I am going to do some LISO modeling to see if the expected electronics noise is a plausible culprit given our current configuration.
A couple of things I notice:
1) If you scan the cavity and look at the relative sizes of carrier and sidebands in transmission, then as you move away from having a zero error signal in the region far from resonance, you can increase the relative size of the sidebands. In other words, the point where we have zero error siganl far from resonance, is also a place where we are not maximizing the modulation depth through the EOM. This is another way of saying we have some offset.
2) If you close the shutter on the laser there is still a 19MHz signal coming from the PD. There is some sort of coupling happnening, and this will give us some offset. I measure roughtly 4mV offset in the error signal with the laser shutter closed.
3) Rotating the QWP also changes the power going into this loop of the gyro since there is a PBS after it. This is a total pain because when you put a PD in and try to reduce the RFAM you are also measuring the reduction in power caused by the PBS. The only way to measure this is to rotate the QWP, measure the power, adjust the power to the same level, then measure the RFAM, and then to iterate.
Attached is a plot of measuring RFAM as a function of polarizer position. Although we moved the EOM, it seems that the position of minimum RFAM is still 100 degrees (actually just a fraction before 100 degrees) where is was possible to get the signal down to approx 450uV (blocking the cavity and using our REFL PD to make the measurement).
Again I saw that this minimum value started to vary a bit, so I set it to the minimum and locked the actuator on the wave plate. Then I recorded the RFAM for a while to see how far it would change. The graph is attached. It varied up to just below 5mV and then started coming back down again.
I measured the modulation depth by sweeping the cavity and using it as an analyzer. The result is arse:
This gives us
Γ = 2 sqrt(10.8 mV / 8.92 V) = 0.07 rads (compare with the 0.25 rads we were getting before).
Also troublesome are the following things:
In a heightened state of despair-induced laziness, I forgot to elog what I did yesterday:
Last week I discovered that there was something wrong with the primary PDH box, so I started there. I was unable to take a transfer function of the overall servo because the output kept railing (I tried adjusting the DC offset). I then noticed that the second switch I had added to optionally remove the DC-gain-limiting resistor in the first, initially non-switchable TF stage (U7) had come loose. I had put this in so that we could get 1/f2 at low frequency, since the baseline design of the servo only allows 1/f. I took a transfer function of this stage alone, and it looked awful. I removed the extra switch and put back the 1k SMT resistor to limit the DC gain, and this made the TF look like it should (a pole and a zero). I am guessing that the OP27 just couldn't handle what we were trying to do with it.
I reinstalled the servo and the cavity wouldn't lock. This time, though, I was able to get it to lock by simply adding flat gain with an SR560 between the demod electronics and the servo input. I concluded that we just didn't have enough gain now that the optical power has been lowered by >20 dB. I needed G = 500 on the SR560, so I changed R23 (between the two TF stages U7 & U4) from 110k to 220, adding roughly the right amount. Note that this is weird, since we didn't lower the optical power by this much... something is fishy.
I measured the OLTF of the primary loop to ensure that we were back at roughly the standard operating configuration. I found that the UGF was at ~10 kHz, which is just about right (the servo gain could probably be increased slightly without oscillation).
With everything running again, I locked the second direction so that I could measure the gyro noise. The OL gain of the secondary loop is now much lower than it was---since I haven't added servo gain---but the UGF is >100 Hz, so we can trust the LF spectrum we get.
The gyro noise now exhibits a strange sort of f-1/2 noise at lower frequencies, and is considerably higher (~10-100x) than it was before we moved the FIs. I suspect that this noise is from a different source than the old LF "hump", and hopefully that contribution is now lower. The task is to figure out what is causing the new noise. I am suspicious of the amount of gain I had to add to the primary servo, so I think it's wise to go back and check our modulation depth, etc. to ensure something's not screwed up.
As of last night, the gyro has not been locking. I have checked everything I could think of---we are getting transmission peaks, reflection dips, and a reasonable error signal, but the cavity won't lock. I finally checked the PDH servo itself by using an SR560 to lock the cavity. This worked, so I'm thinking there is something wrong with the servo. It must be early in the chain, because I am able to generate signals at the output with the sweep input without a problem.
Last night I calculated new modematching solutions for both beams using the profiles measured yesterday morning. The REFL beams now return to the FIs at a manageably small size (very close to the size of the beams coming through from the input side, as they should). See the picture below. Note that both beams are incident on the card: input from one side and REFL from the other. It was impossible to notice a difference in size by eye---this is good.
I went about realigning the cavity mirrors over again, since it seems like they might have been bumped around with all the commotion. I then locked the cavity in the CCW direction and fine-tuned to get about 50% transmission. We should be able to get about 90% with reasonable losses and our R values, but we've never seemed to be able to get much higher than half. The modematching solutions I calculated should give us the standard >99% coupling via the naïve model, but I still maintain that we can't get a truly flat phasefront with an astigmatic input beam, and can therefore not get the coupling we calculate with 'EtaCalcSimple'.
I was nearly done with aligning the CW beam when something happened and the CCW refused to stay locked. Things were also difficult to work with because the output optics were no longer aligned so that the two output beams were overlapping, so I couldn't simultaneously monitor both while aligning stuff. The best way to do this is to reconfigure all that output stuff the way we used to have it and then go about optimizing both beam alignments. Working on this today.
Here are the new beam profiles after the faraday isolators. I am satisfied with the CCW profile, but the CW looks a bit crap. It is much more astigmatic, and I'm afraid that some diffractive effects near the output of the FI are tricking the fitting algorithm. We are going to re-profile the beam after the MMT, so we will verify that it is matched to the cavity mode and make adjustments if necessary.
It seems like the consensus is that the best idea is to place the faraday isolators before the MMT. This way, we can use independent lenses to focus the beams into the FIs to avoid clipping and not have to deal with this additional constraint when designing the MMT and when adjusting the steering mirrors.
We played around with the layout this afternoon, and there just isn't enough room to fit both FIs and PDs (along with steering mirrors and lenses) side by side far enough upstream to leave room for modematching. We decided to rotate the AOM setup so that it folds back towards the laser instead of perpendicularly outward, as in the picture below.
The second picture shows that the FIs and PDs fit just fine with this newfound space.
Our current idea is that clipping on the faraday isolator is resulting in disturbance of the error signal. The precise mechanism isn't quite understood yet, but there are several candidate theories. We have moved the FIs so that they are at the waists between the lenses of the mode-matching telescopes, and the beams going through them from the input sides (from the laser) are small enough that no clipping is evident as measured with a power meter.
The same can't be said of the beams returning from the cavity mirror. Moving the FIs to the MMT waists has highlighted the fact that the REFL beams do not seem to be of the same size as the input beams at this point, which indicates a problem with mode-matching.
Small changes to the MMT lens locations don't seem to do the trick, and since this is such a crucial aspect of our experiment (given the isolation method), it will probably pay to just redo the thing. Alastair had the idea of profiling the beam closer to where the lenses actually are/will be---instead of as farther upstream where we have done it in the past---and that way we can rule out anything weird going on in between. Since we have a fair amount of space at the right place on the table, we can also profile the beam after the lenses are in place by steering the beam to the unused area. This way we can actually verify that the beams have the right shape on the way into the cavity.
I moved the faraday for the CCW return path back into the middle of the MM telescope, where the waist is. It turned out that the CW PD interfered with this, so I had to also move this back to the waist of the other MM telescope. The beam wasn't quite at 4" through the telescope so I had to alter the height, which in turn un-aligned the beam for the CW input path.
I was able to get it looking like the input beam and the reflection beam were not clipping on the outside of the faraday. Measuring the power that was being transmitted through the faraday into the cavity there was no measurable loss. However, when I looked at the beam rejected out the side (going to the PD) it was very weak. I found that you could increase this by twisting the faraday slightly, however even then I was only able to get 1.26mW out of 1.56mW that was going in. This is a serious disadvantage to this Faraday design!!! You cannot see where any clipping is happening internally, even when the beam in is small at both sides, and there is no clipping on the outer apeture. It would make life a lot easier if the input and output polarizers were separate.
Next I started aligning the beam back into the cavity again. Of course we haven't touched the cavity mirrors so that is still aligned. The CWW input beam had also become misaligned, though I have no explantion for this. There were a few mirrors with screws that were not very tight, and I tightened them down. Afterwards the CWW input beam was misaligned, so either something moved as I tightened it or I bumped an actuator. After getting the beam vaguely aligned I then looked back at the faraday again. Of course the return beam is now misaligned.... now I can't get more than 50% of the beam back through without clipping.
I think that this is going to require some iteration. We'll need to just keep trying to keep the faraday aligned as we get the beam back into the cavity.
More mis-alignment tests to start with today.
I moved the faraday for the return CCW beam (the ingoing CW beam) an un-calibrated-small-amount to the side (about 1mm). The transmission through the cavity gets worse (from 720mV down to 400mV DC on the trans PD). The noise hump at low frequency gets worse by a factor of 2.
This test is really a precursor to trying to move the faraday back to where the waist is in the mode matching telescope in order to reduce the clipping.
Since we are wondering what effect pointing may have (in particular with respect to coupling into the cavity and through faradays and onto PDs etc) we decided to do a little deliberate misalignment to see if the noise was impacted at all.
Firstly we misaligned the beam into the cavity. We used only the horizontal actuator on the last mirror before the cavity for the CCW beam. Looking at the trans PD DC voltage we moved from the initial max transmission (with the plastic cover in place over the 2 input mirrors) to lower transmission by moving the actuator CW and then CCW.
The result is that at lower frequency the noise was slightly improved with misalignment. The misalignmend in the CW and CCW directions produced strangely similar results in that we get a factor of a couple at low freq. At frequencies around 1Hz and above the noise gets worse due to the box being removed. We confirmed this was the cause by going back to the max transmission (now only 740mV) with the box off and seeing that the higher value stayed at around 1Hz. Interestingly after this moving around the noise at max transmission was now a little better than when we started.
Next we put the mirror back to the max transmission, and put foil over the small path between box and vacuum system which was worrying Koji. You can see the effect most clearly in the second pdf below. The noise between 0.4Hz and 6Hz is improved by a reasonable amount.
The final test we did was to misalign the beam onto the CCW PD by a small amount. It would have been nice to quantify this and to know where to put the beam back to, but the DC output is so noisy (really terrible) that it is just impossible. I misaligned it gradually using the horizontal actuator on the mirror right before the lens that focuses the beam onto the PD. I moved it until the cavity would not lock, and then came back just a little until it would lock again. It is much further toward the edge of the PD now, though it was difficult to tell whether any of the beam was clipping on thed edge (at least I couldn't see any through the viewer). As you can see in the last pdf below the spectrum is quite a bit worse.
So while this doesn't give us the answer, it perhaps gives some clues as to what is going on here.
I finished up by trying to realign the beam on the CCW PD. Since the DC level is messy it is difficult to know what to use for this. I used the viewer, and then swept the cavity and got the error signal up on the scope. It seemed like the beam was already giving the full error signal, and that trying to move it around using the mirror before the lens didn't give any difference to the height of the error signal until the beam got right off the edge of the PD (around 7 full turns of the actuator in either direction). One last FFT showed that the noise is back down to the level it was at before I misaligned the beam on the PD.
PS: I also notice that we don't have the vacuum system pumped down at the moment. Clearly it's not making a great deal of difference, but I guess we should pump it back down again soon.
For lack of a great way to look at the low-frequency noise in the REFL PDs (due to the noisy DC outputs described earlier), I went back to the old idea of using a pickoff from near the laser output to look at intensity noise. The setup is a little bit hacky, since I just shot the beam through the side of the box with no viewport, but I think it's fine for this measurement. I took the light that was being dumped from the initial PBS. It was strong enough that even through the box I still had to attenuate it using an OD(=2) filter.
The spectrum I saw had increasing noise at lower frequencies, as should be expected I suppose. It didn't look quite right to explain what we see in the gyro signal, but I it wasn't obvious enough to rule out. I built a simple ISS servo by feeding it into the digital system, subtracting an offset, putting it through a 100-Hz LPF and feeding the output to the power adjust on the laser controller. I was able to get ~200-300x suppression below 100 Hz without the loop becoming unstable. The loop had essentially no effect on either the gyro noise or the TRANS_DC spectrum, leading me to conclude that the excess noise is not from input power fluctuations. I guess this is somewhat of a relief, as I have no idea how that would cause the noise we're seeing anyway.
Here is a before/after plot of three signals: the gyro noise, the TRANS_DC signal, and the ISS PD signal. NOTE: The ISS PD signal units are arbitrary, so the fact that the open-loop noise is right with the other traces around a few Hz is meaningless. As in the last few entries, I have multiplied the TRANS_DC spectrum by a phenomenological factor of 10 to show that it has precisely the same shape as the gyro noise at low frequencies (that is less convincing in this plot due to less averaging and higher FFT BW, but look to old posts to be made a believer).
I did the same comparison as before using data from today, and the results are a little weird. The gyro noise is the same at low frequencies, and the DC_TRANS spectrum (when multiplied by 10) is also coincident at low frequencies. However, the high-frequency floor of the DC_TRANS spectrum today is higher than it was last night, and there is also a little bit of excess HF noise in the gyro spectrum. I am going to check what the dark noise of the TRANS PD is now.
I also took spectra of the DC outputs of the REFL PDs, and they don't share this low-frequency bump in the spectrum with the TRANS and gyro signals. This is above the ADC noise floor, but it could easily be the electronic noise of the DC output of the PD (which is very noisy for some reason). The point of this measurement was to verify that we saw power fluctuations on the REFL side in coincidence with the TRANS ones. Then, we would see if they were in or out of phase with the TRANS fluctuations to see if they were input power fluctuations or cavity coupling fluctuations. This would essentially give us the information we would get from setting up a pickoff PD near the laser to monitor input power fluctuations, plus a little more.
I will see if this is dark noise and then figure out a way to make a more precise measurement of the REFL fluctuations.
I confirmed with Hiro my suspicion that the situation described in the quote probably can't really be what's going on. As in the case of a regular fabry-perot cavity, the mirrors themselves define the mode and they must move for any rotation or translation of the mode to take place. So, pointing drift in the input beam can't cause a differential path length between the CCW and CW modes, though it would cause us to see AM at the output (like we do) as the coupling would go up and down.
I guess it's possible that the mirrors themselves can be twisting and causing the mode to wander, but I think geometry dictates that the effect would be the same for both directions.
I can't say that I can think of any obvious coupling mechanism here. I think it's feasible that pointing drift of the input beam is causing the spatial eigenmodes of the cavity to wander differentially (i.e., the points at which the supported mode in one direction touches the mirrors all move horizontally across the mirrors a bit, making the whole square 'rotate'---this is dependent on the input beam and can thus happen independently in each direction). In this case, the cavity length would differ in the two directions, resulting in a gyro signal.
That is for some near-DC AM to be converted to AM of the 19-MHz beat frequency. The idea before was that jitter at 19 MHz would turn into a 19-MHz signal from misalignment into the AOM or via any other mechanism that turns jitter->AM (e.g., clipping). This looks like a DC error offset. Then, any low-frequency noise (e.g., slow pointing drift of the steering mirror on the way into the AOM, slow translation of some aperture relative to the beam) causes low-frequency AM of the 19 MHz signal, which looks like a time-varying error signal (noise). AM of the AM.
So, there is some 19 MHz oscillation set up by jitter->clipping or polarization->AM or some gremlin waving his hand steadily across the beam coming out of the laser at 19 MHz. Then, the amplitude envelope of this oscillation is modulated by some low-frequency noise source, and this is what looks like noise when demodulated by the PDH setup.
The question I'm wondering about now has nothing to do with envelope modulation of a 19 MHz signal. Instead, there could be low-frequency pointing noise on the way into the cavity. The cavity reflectivity looks like a real number close to 1 for the sidebands, independent of small angular misalignments. So, any existing 19-MHz signal is not amplitude modulated. The question is: how does this sort of thing become noise? I thought perhaps the eignenmodes of the cavity rotate with respect to each other since the beams are not being injected from symmetric points, and this could cause some relative length change that looks BIG in the gyro signal compared to what a common-mode length change would look like via FSR modulation. This could be why we don't see it in the primary (laser) actuation signal.
I think you raised the possibility that the same low-frequency pointing that causes us to see noise in the DC_TRANS signal could also produce RFAM via clipping or whatever, but in this case we would still see the noise with the cavity obstructed, which we do not.
In terms of a coupling mechanism, I'm wondering about this: If the pointing of the carrier as it goes into the cavity is moving around at low frequency, then that's going to modulate the coupling into the cavity and give us some AM on the reflection PD (I'm not talking about RFAM here, but low frequency AM as the coupling to the cavity wanders around). This then causes the AOM to try to act, so we see noise superimposed on the reflection and transmitted beam.
It seems that whatever the cause, if we are seeing AM at low frequency in the transmitted beam then it is likely that we will also be seeing the exact same signal on our REFL PDs.
EDIT: Zach pointed out to me that there needs to also be some phase shift between the carrier and sidebands for us to get an errors signal. I think that I'm not clear now how we ever expected the RFAM to show up as noise. Where is the phase shift there?
Since it looks like AM at the input side is not the root of our current low-frequency problem, we have been trying yet again to come up with new sources. Alastair had the idea that pointing noise of the beam going into the cavity might be causing us a problem in some not-yet-well-understood way (this was first discussed in the context of 19 MHz jitter from the EOM being converted to RFAM as it could be in the AOM, but the cavity should have very little response at 19 MHz with which to do this).
I decided to look at the low-frequency spectrum of the DC_TRANS signal to see if there was the same type of behavior as in the gyro spectrum. As it turns out, there is. Below is a plot of the two compared side-by-side (though I had to scale the DC_TRANS plot up by a factor of ~10 to get it to coincide). There is extremely good agreement in the shape of the curves, and I think we'll see that the HF floor of the DC_TRANS plot is just the noise floor of the (broadband wide-area Thorlabs) PD.
We need to do some more work to figure out what exactly is going on, but this is another data point that helps with the diagnosis. I think the next step is to look at the power at some pickoff point close to the laser to see if this is an input power drift or something caused by varying degree of coupling into the cavity (as from misalignment).