Really cool plots, though I'd prefer the scales for a given octet to be frozen so as to not hide variations in magnitude. I feel like we're finally starting to see the whole elephant rather than just feeling the ears one week, the tail the next week, etc.
When comparing even and odd octants, note the even octants contain bare bars which are ~20x more sensitive to gamma rays coming from the front end. So to explain the correlation Jim noted in octants 2 and 4 in hclog 201130, the throat of the Pb collar could be a little brighter on beam left than on the beam right side. From Katherine's simulations, I have the impression all it would take would be a 1-2mm offset between the center of the W plug and the center of the Pb collar. We may already have the data to test that hypothesis from the recent +-1mm beam steering runs.
BPM noise: I'm not sure how significant it is to remind everyone that the S/N ratio in the BPMs is inevitably much worse than in the old 30 Hz measurements: the electronic noise is roughly 200 microns, much larger than any expected beam jitter (even the 60 Hz is only of order 100 microns, and FR averaging fluctuations are smaller still). So expected position correlations in plots of natural beam motion should be subtle, a fraction of the electronic noise widths. For the most part that's true.
But something is broken: The clear Y correlation Katherine sees is a big surprise, yet the the up-down anti-correlations prove it's real. The magnitude is not only beyond expectations for power line harmonics, it's a complete shock to see that much motion in the vertical direction. (Note that the magnitude seen here is a lower bound: our parity DAQ suppresses 60 Hz differences a lot but 600 Hz hardly at all.) Assuming our own BMod was off during run, then we've got a vertical corrector with a lot of ripple on it and effectively no working FFB in the vertical direction. If it's created in the 3C line, one could track it down by looking at the position differences on the way to the target. Paul King says the FFB is currently "on". I think this is evidence the FFB is not fully operational yet. It would be nice to get it working this quarter.
If you take a knife to a gunfight, make sure it's sharp: Since the average beam position isn't moving several mm in this run, the x*dx and y*dy correlations are presumably specious. You can't distinguish X*dX correlations from dX correlations without really swinging X: otherwise the dataset is basically constant*dX. Those correlations are expected to be 1-2 orders of magnitude smaller than the 1st order correlations anyway, so they probably won't be visible/fittable except in the averaged asymmetry of opposite pairs of octants.
references:
BPM noise level is based on a 6 micron/sqrt(Hz) benchmark I inferred from Riad's 30 Hz injector measurements several years ago.
In Bscope measurements from 1997 probably predating FFB, the amplitude at 60 Hz was O(100 microns). I think a properly working FFB can squash that down to 25 microns but I couldn't find any examples in the hclog. Here are some older examples:
https://hallcweb.jlab.org/hclog/9712_archive/971219110859.html
https://hallcweb.jlab.org/hclog/9711_archive/971124151507.html