40m QIL Cryo_Lab CTN SUS_Lab TCS_Lab OMC_Lab CRIME_Lab FEA ENG_Labs OptContFac Mariner WBEEShop
  PSL  Not logged in ELOG logo
Entry  Thu Aug 12 17:34:37 2010, Megan, Summary, NoiseBudget, Noise Budget NoiseBudget.figNoiseBudget.png
    Reply  Fri Aug 13 15:10:33 2010, rana, Summary, NoiseBudget, Noise Budget 
       Reply  Fri Aug 13 15:47:36 2010, Megan, Summary, NoiseBudget, Noise Budget BeatSignal.pngNoiseBudget.pngNoiseBudget.fig
Message ID: 280     Entry time: Fri Aug 13 15:47:36 2010     In reply to: 279
Author: Megan 
Type: Summary 
Category: NoiseBudget 
Subject: Noise Budget 



Sorry, here's a prettier graph! I think the strange bumps in the low frequencies is from very few samples in that region and the interpolation matlab is trying to do. Once I have more data to compare it to, I'll try to figure out how to get rid of them.

 It looks a little like the FFT was done wrong here. How about uploading some more details, e.g. what is the channel names used? Is this in the DAQ or just using the SR785? How about a diagram showing what the transmission PD is, what the range of the Marconi is, how the signal goes to the DAQ, etc, etc, etc.....

Don't we need a whitening filter to put the PLL control signal into the DAQ? Or I guess we can use a DC channel and an AC coupled, high gain channel.

 This was directly from the SR785, where I saved the data using the GPIB scripts. This is the data with the 1MHz input range of the Marconi with the New Focus photodiode signal. The two signals go into a mixer, through a 1.9MHz LPF, then to the Stanford SR560 with a gain of 2000 and a break frequency of 0.03Hz on the low pass filter (the values necessary for it to actually lock). This feeds back to the input of the Marconi. We looked at data from the error point and the feedback loop with an input range of 100kHz to compare, calibrating the error point by pi/Vp-p*UGF (compensated for the gain by multiplying by 1/f, but then also multiplied by f to give the frequency noise) where Vp-p was 836mV and UGF was 16.2kHz as measured by taking a transfer function with the SR785. The feedback loop was calibrated by multiplying by 71kHz/V, the measured range of the Marconi (measured by applying 0V and 1V and finding the frequency difference). The 1MHz data was from the feedback loop and calibrated by multiplying by 715kHz/V (measured range at 1MHz setting). All the curves showed the same features and were at roughly the same level (within a factor of 2). I chose to plot the 1MHz data because it extended to lower frequencies (possibly a poor choice). We could not get lock with 100kHz input range for long enough to measure the lower frequencies because the cavities were drifting too much in temperature. We looked at the channels for temperature, and it looks like a 1K step in room temperature was causing the cavities to drift quite a bit. So our goal is to restabilize the cavities and take another measurement to verify the ones we already took.

*EDIT* Frank pointed out a concern with the lower frequencies. I attached a graph with the actual data. The strange spikes in low frequencies is from the interpolation of data that isn't real to begin with because of the measurement process. I'm working to figure out how to fix the interpolation and will hopefully have a graph of something less confusing soon. Also, the flat line at higher frequencies is electronic noise, we believe from the demodulation setup, because the line is always present, but higher with higher input ranges.

*EDIT2* The noise budget is back! I just deleted the low frequencies that aren't accurate and the high frequencies that weren't measured because I haven't figured out how to accurately extrapolate what we have.

Attachment 1: BeatSignal.png  13 kB  Uploaded Fri Aug 13 17:13:46 2010  | Hide | Hide all
Attachment 2: NoiseBudget.png  51 kB  Uploaded Fri Aug 13 17:36:51 2010  | Hide | Hide all
Attachment 3: NoiseBudget.fig  122 kB  Uploaded Fri Aug 13 17:37:01 2010
ELOG V3.1.3-