Bots, Bureaucrats, Interface administrators, Suppressors, Administrators, Upload Wizard campaign editors
573
edits
David Notebook: Difference between revisions
Jump to navigation
Jump to search
no edit summary
No edit summary |
|||
|
==Daily Log==
'''
The eBay mirrors ended up working. I'm using two curved ones with R=75 cm, so the waist is in the middle of the cavity now. The size was only about 10 microns different, so I basically just moved the cavity forward 5 cm, and what do you know--huge transmitted and reflected peaks. I'm guessing it's better because the mirrors must be very close in reflectivity, but it's weird because the impedance matching equations (see the python script) suggest that it shouldn't be a huge deal assuming the reflectivities differ by a couple tenths of a percent (seems likely for any cavity mirrors). Maybe there was something about the plano-concave configuration? We don't quite understand something here, but sometimes you just gotta move on.
With a bigger reflected signal it wasn't too hard to get the laser locking to the cavity. It was easier to lock to non 0,0 peaks. Even though the signal was smaller than the 0,0 mode ones, they were more stable in transmitted power, which made it easier to lock to and the lock more robust. I don't want to send to much power to this cavity though since we need most for the fiber amplifier, so for now I'm locking to the 0,0 mode but I might switch back. The cavity seems much more responsive than before. When I adjust the MML, I can see the cleanness of a FSR change. It looks about as good as the Raman cavity.
The lock still isn't perfect, but it's much better than the Raman lock--I've gotten a lock maintained for ~1 hour, which is already good enough. I think I can probably still improve it. In addition to the locking circuits, putting the cavity in a padded tube might help damp low frequency noise which seems to be the dominant error signal frequencies.
So now is the hard part of the whole locking 1 laser to 2 cavities thing. There's a lot of ways I can see to do this. We generate a second error signal from the Raman cavity--we can try sending this to a lockbox and feeding back just to the cavity piezo (maybe the laser line is sufficiently narrowed from the low-finesse cavity). We could send both error signals to one lockbox (it turns out you can add two signals just be using a splitter backwards! Who knew? No one in lab, that's for sure). But that could be tricky to try to get it to lock to two things at once, even though the frequencies are pretty different. After locking one, trying to adjust the DC offset to get the other signal locked would probably disrupt the first lock. I could probably lock the low-finesse cavity with just the fast feedback and then try using the aux servo output on the same vescent box to lock the Raman cavity--this has the same issue though of having to adjust the DC offset, but would have less competition with the fast feedback.
What I think is most likely to work is to use a second lockbox and feedback either directly to the laser diode or combine error signals and send both to the current driver (I don't see any advantage of feeding to the diode, and that would be difficult/invasive to setup). I tried the combining error signals and feeding back to the current driver (slow-feedback to the separate piezos today). One problem I had was with the RF signals. At first I used a second RF generator for the Raman error signal, but I noticed a lot of noise. I eventually realized it was at 5 Hz and tracked it down to the fact that the RF generators aren't super accurate and the signals differ by about 5 Hz even when they are both set to exactly 50 MHz. The problem got a lot better when I reduced the frequency of 1 by 5 Hz, but it was still a little off and I didn't have the resolution to improve it more. So it seems like all the signals need to come from one RF generator. But without some crazy combination of splitters, I can't send the same power to all components. '''I eventually upped the power of the 1064 RF generator to 13 dBm''' (this is now the normal power that should be run at), split the output to send 10 to the Raman error signal, which leaves 10 to be split into two 7 dBm signals for the EOM and low finesse locking signal as usual. But this means that the Raman cavity error signal is getting 10 dBm instead of the usual 7. It probably doesn't really matter, but this is different than before. Maybe I can find an attenuator that'll work?
Anyway, once that was all worked out I wanted to try the double locking thing. In short, it didn't work yet but it seems like it could be possible--I had two error signals going into the current driver and nothing crazy happened and both Vescent boxes seemed like they were almost locking. I was having trouble with the low-finesse lock, which I think it just because the RF power is probably slightly different, and more importantly the summing signal thing (backwards splitter) probably introduces a phase delay that I haven't properly accounted for. I want to check with RF or function generators tomorrow and see if I can figure out how much it delays things. Hopefully there's no dispersion because I don't know how I could fix that.
It would also be good to check the linewidth when locking to the low-finesse cavity once new Zach gets the interferometer running.
'''1/22/15'''
Still haven't heard back from Layertec. Thorlabs says they can do it, but it'll be ~7 weeks, so I'm holding off on that for now. Deniz says Lambda Research is usually pretty fast. I requested a quote from them but haven't heard back yet. Zach found some mirrors on eBay that don't give a ton of info, but seem like they might work. I ordered them and hopefully with that we can make something happen.
'''1/20/
I put an arrow on the backpolished E03 mirror pointing towards the side that was facing upwards in the packaging. I think this is the "back" side that is less well polished. The mirror didn't seem to help the signal strength though, and in fact made it worse. I flipped the mirror around too in case I got it in backwards, but I didn't notice any change. The problem now is probably that it is poorly impedance matched because the reflectivities are so different, which apparently is a thing. I updated that python program to account for this--we're probably only getting a few percent of the full height of the reflected signal for resonance dips just because of the impedance, so that's probably the issue now. I'm seeing if Thorlab can make us a custom curved mirror (2 E03s would give a finesse of ~800) and I'm talking to Layertec about making a pair in the 2000-3000 range. We want the highest finesse we can get away with that will still be easy to lock and stay locked, but we're not sure what value that will end up being.
| |||