Investigating cavity ringdown measurement

From the last comment on https://mccullerlab.com/logs/lab/index.php?callRep=11650 we can clearly see some additional process decaying at the tail end of the cavity ringdown measurement. Attached is a screenshot of a quick check I performed where I repeat the cavity ringdown measurement process with no cavity. This is pure reflection from the input coupler. Attached is a screenshot of triggering on the decay. There are three distinct time frames:

1) -375.4 ns to -294.2 ns - This is the first moment where the total light on the PD is changing. Note that this portion is flatter than the following one. Probably some initial slow process on the AOM. This looks like the AOM is slowly becoming less efficient immediately after the power is cut.

2) -294.2 to -234.2 ns - Rapid decay of light on the photo detector. This could be the light travel time? 60 nanoseconds is order of magnitude how long the light is taking to travel to the PD (don't know the exact refractive index of the fiber).

3) -234.2 ns and on - This is the same shape as the tail end of the cavity ringdown measurement. This seems like some residual voltage produced by the PD on short timescales.

Also note that we don't see the large blip that is in every cavity ringdown measurement perhaps suggesting it is not from the AOM but some interaction with the cavity.

Syncing Moku Clocks

[Torrey, Daniel]

I connected the 10 MHz Reference In on Moku 1 to 10 MHz Reference Out on Moku 4 with a bnc (not a homemade additional shielded cable). When turning on the EOM drives to the 775 nm light EOM and the 1550 nm light EOM, one from each iPad, we don't observe a sine wave modulation around 11 Hz like we did in this post. If we shift the frequency of the EOM drive on one Moku, we can see the sine wave.

The Moku Website says to "ensure 'always use internal reference' is disabled", but we did not find this setting. The third (right-most) LED on Moku 1 indicating there is a reference input is blue. Considering this LED is lit up and the 11 Hz sine wave went away but is recoverable by intentionally changing the frequencies away from each other, I am assuming this setting does not need to be changed.

Optical Layout proposal for new lab space

Here's a proposal for a layout for the first parts of the experiment in the new lab space (attached). The plan was to have space for an in-air prototype Michelson, and simultaneously have space for building up a power-recycled IFO, with both connectable to the photon counting readout, and hopefully minimise movements of components in moving to later stages of the experiment. The power-distribution, SHG, and input laser breadboards are intended to go either on shelves above the benches or a bench in the other lab. Layouts within individual breadboards are rough drafts. Let me know what you think.

Cavity transfer function measurement and system identification

In preparation for an adaptive scheme for filter cavity controls using buzz, I've written a jupyter notebook to lock the cavity, take a transfer function, and perform system identification using tools from wield.

Attatched are some plots of the cavity open loop transfer function, the controller, and the plant model with different scan amplitudes. They all used the 'dynamic amplitude' mode on the moku

Effects from Bit Resolution on Biquad

I attached four plots characterizing the precision of the transfer function for different bit resolutions (assuming a 10MHz sampling rate and a critical frequency of 10Hz). In the meeting yesterday, Lee mentioned how the lines were not normalized, so I updated the graph to have normalized lines (Figure 1). However, in that graph, all the lines (excluding the 18 bit one) are aligned perfectly which is odd and still something I am trying to understand. Regardless, I included the non-normalized version for reference (Figure 2). I also included a graph showing the ratio between the float version and truncated versions to see if there were noticeable deviations due to different bit resolution, but there were not (Figure 3). For the remaining graph, I changed the bit resolution for all of the coefficients rather than just the b coefficients (Figure 4) to explore what that would look like.

Finesse via a cavity ringdown measurement

[Torrey, Ian]

We had attempted to make a cavity ringdown measurement in the past unsuccessfully (never posted on the log). Here we show a successful measurement. Key difference to past attempts were the TRANS pd we originally used were low bandwidth (BW<550kHz) where the time scales being measured are on the order of megahertz. Previously we also used the FSR of the 1550 light instead of the 775 light, which is just a factor of 2.

Here is the new attempt. Changed the measurement PD to the 1811 REFL PD to DC coupling and adjust error signal set point and PID controller so that you can lock the cavity with the much higher bandwidth PD on DC coupling (DC coupling the PD reduces the error amplitude). We lock the cavity with the 775 light using this method and then change the trigger mode on the moku to decaying edge and set the threshold underneath the lock point. We then use the AOM as a "fast shutter" of the light going to the cavity simply by turning off the amplified RF drive going to the AOM. This should be a sufficiently fast "blocking" of the light going to the filter cavity. We see a decay.png in the cavity by doing this. I then wrote a quick script to fit a decaying exponential to the data and calculate the finesse from this where the finesse will be F = 2*pi*FSR*tau = 2*pi*(3e8/2.4)/2*tau (from here tau * cavity bandwidth = 1/(2*pi)).

How good the fit is depends largly on the portion of the data used, where the tailing end of it does not seem to be exponential. For example:

If we use this much of the data decay-overlay.png, the fit looks like decay-fit.png (decay-overlay.png). This yields a finesse of 389.

If we use only the portion of the data that looks like an exponential (the end of it looks linear,not sure why) decay-data-used2.png, the fit is much better decay-fit2.png/decay-overlay2.png and we get a finesse of 303.

The finesse should be 313, see post [11633], based on the reflectivities of the mirrors. Further analysis is needed but I think we can constrain the measured finesse to between 389-303.

 

Morning work 6/11

[Daniel, Torrey]

Talking with Sander this morning he pointed out that the amount of power on the 1550 camera from both beams should be approximately the same. Attached pics with the camera show that while locked on 775 and attenuating 1550 light, and blocking the 1550 beam, the majority of light is 775 leaking through. This was done for 2.2 mW input power for 1550. I have increased the power to 15.7 mW input for 1550 so that the amount of leakage of 775 through the mirror will be negligible.

Below are a brief description of each image.

• "1550 blocked low power mode" and "both beams low power mode" are 2 mW input power, unknown exposure time (these are looking for 775 leakage, which they do)
• "lowpower-increased input.bmp" is 15.7 mW input power with an exposure time of 29859 us, gain = 0db.
• "highpower - increased input.bmp" is 15.7mW input power with an exposure time of 16 us, gain = 0db.
• "1550 blocked lowpower-increased input.bmp" is 15.7 mW input power with exposure time of 26163 us, gain = 0db; 1550 blocked
• "highpower - increased input ND 2 filter" is 15.7mW input power with an exposure time of 70 us, gain = 0db; ND 2 filter applied in front of camera (used so that we can fit Gaussian to high power beam without a saturated image)

With these measurements and this script I'm getting 8 * 10^4 suppression. I think the low power measurement method needs to be improved for this to be reliable. 

Fitting a saturated beam with the basler cameras

This is an update to 11401. I have been looking for an accurate way to measure the amount of power exiting the cavity when the cavity is locked with 775 but in the configuration that attenuates 1550 light. This is easy when the beam is wellbehaved.png. However because there is a massive difference in powers we need to measure in the two configurations, we must either turn the power way up so that we can measure the attenuated beam, and struggle with the saturated coresonant beam. Or, turn the power way down and easily measure the coresonant beam but struggle with the attenuated beam. This script should solve the former, where I have applied a mask to the data array that makes up the camera data, getting rid of any saturating data and then fitting to that portion of the data instead, seen here: saturated_solved.png. 

I am still looking into a proper way to sum these values without getting the background light. There may also be some pick up of the 775 light on the 1550 camera in the low power setting as the expected amount of light from both should be order nanowatts. I am going to retake these pictures (and remember to write down the exposures) now and will update.

PDH lock with Moku API

The attached Jupyter notebook will do the following:

-Set up our 'standard' multiinstument setup, with laser lock box, frequency response analyzer, digital filter box, and spectrum analyzer

-Configure the digital filter box to perform a sum and all-pass filter

-Configure the laser lock box to perform a sweep of the laser frequency and record the transmission and PDH error signal

- Find the voltage at which the transmission is maximized

-Set the offset to that voltage

-turn off the scan

-turn on the controller

It locks! You may have to run the bottom of the script repeatedly in order to aquire lock, but it hits eventually. Lock is lost if you attempt the handover to the iPad, but can be quickly reacquired.

 

Input/transmission/reflection measurements of filter cavity on resonance

[Daniel, Sander, Torrey]

We have taken input, trans, and reflected powers for the cavity at both wavelengths. This is done by locking the cavity with one wavelength and then measuring the respective powers of the other. Additionally when doing the 1550 nm light measurements you have to tune the AOM RF drive frequency so that the 1550 transmission/reflection is maximized/minimized. The results are as follows:

1550 nm

• input - 2.18 mW
• measured output .67 mW (this is after a 5050 BS)
• Infered output ~1.4 mW (.67 *2 + 5% losses on the BS)
• reflection is .5 mW
• This implies a 13% loss.

775 nm

• input 94 uW
• transmission 64 uW
• reflection 46 uW - this requires swapping the locking scheme to be on the 1550 path as blocking the 775 refl path would obviously break the cavity.
• Parking the two transmission peaks is more difficult in this configuration and is providing numbers where trans + refl > input which don't make sense. More to come on this measurement.

Filter cavity suppression crude measurement

[Daniel, Sander, Torrey]

Attemped to measure the amount of power suppression the cavity is providing for the 1550 path. Plan was to use a camera, take a pic, fit a gaussian to the beam, then integrate to get total power. Do this for both beams without having to calibrate into units of power we could get a relative suppression via the ratio. THe problem is there is too much power when the modes align, as the camera saturates. And the 1550 ND filters we have are dirty. I attempted it with the power meter and got the following transmission values:

.550 mW When the two modes are maximally coresonant.

9 nW When the two modes are minimally coresonant. (blocking the 1550 beam reduces the reading down to the background level)

7 nW Background.

If we believe this very crude measurement this is a suppression factor of 2.75 * 10^5.