Compton Cavity State Identification (d2n)

Background

In the Hall A Compton Polarimeter, Compton scattering occurs inside a Fabry-Perot cavity fed by an infrared (1064 nm) laser. Over the course of a run, the cavity is periodically filled with photons of a given polarization state (right or left circular), allowed to empty, and filled with photons of the opposite polarization state.

In order to compute accurate Compton asymmetries, we need a reliable, accurate way of determining the state of the Compton laser cavity during any given mps:

• Is the laser on and locked to the cavity, or unlocked? This is the difference between a few milliWatts of power and 400+ Watts.
• Are the photons in the cavity left-circularly or right-circularly polarized?

Cavity State Variables

During d2n running, two types of signals were used to keep track of cavity state data: "real-time" bits read in through the TIR module, and EPICS variables (which can be much larger), which are read into the datastream roughly every 50 MPSes. One obvious drawback of EPICS data is that, due to its less frequent readout, it lags actual changes.

The variables at our disposal are:

• Cavity power
  • A "real-time" bit that is supposed to read LOW when the cavity is empty, and HIGH when the cavity is filled. This bit is buggy.
  • A floating-point EPICS variable that is supposed to report the actual cavity power in Watts.
• Cavity photon polarization direction
  • A "real-time" bit that is LOW when the photons are left-circularly polarized, and HIGH when the photons are right-circularly polarized.
  • An EPICS variable that reads -1 when the photons are right-circularly polarized, and +1 when the photons are right-circularly polarized. (Note that these two variables use opposite conventions.)
• Cavity photon polarization
  • An EPICS variable giving the result of a polarization measurement on the photons exiting the cavity, in percentage points (e.g. a reading of 94.1 corresponds to 94.1% polarization).

Cavity Power Calibration

Originally, the analysis code for the CMU DAQ decided that the cavity was ON when the following two criteria were satisfied:

• EPICS cavity power reads greater than 300 Watts
• Real-time cavity power bit is HIGH.

However, we noticed a problem when we separated the trigger rates according to the on/off status of the cavity. As you can see in Figure 1,

Fig. 1: Trigger rates from CMU DAQ for cavity on/off

the "cavity on" data (blue line) displays a secondary peak at the precise location of the "cavity off" rates. This strongly implies that some events where the cavity is empty are being misidentified as belonging to a cavity-on state.

This misidentified subset of events was resistant to time-based cuts (e.g. excluding the first ten seconds after a laser-state transition). We decided to examine a small number of laser cycles to see whether there was a pattern to the misidentified segment. Fig. 2 shows a strip chart of the trigger rates for the first ~17 minutes of a run. Blue dots have a "cavity ON" label; red dots have a "cavity OFF" label; and green dots have a "cavity UNKNOWN" label (that is, the variables disagree). The green events near the beginning of each high-rate (therefore laser ON) state are expected due to the lag in the EPICS variable turning on. What's unexpected are (1) the blue events near the beginning of each low-rate (therefore laser OFF) state (we expect these to be green) and (2) the green, or unknown, bands in the middle of each low-rate state. (The drop in rate around MPS=21000 is due to a beam trip.)

Fig. 2: Trigger-rate strip chart from Run 60732, split by identified laser state

Fig. 3 shows the same strip chart, this one with the two cavity power variables superimposed on the chart for comparison with the rates. (Each has been multiplied by a constant factor -- 13 for EPICS and 4500 for real-time -- to make it visible on the display. Zero power will still read at 0.) You can see that the violet (EPICS) readouts behave as expected, trailing the actual cavity power slightly. The black (real-time) readouts, however, do not look so good. Although they come on promptly when the laser cavity does, they often lag significantly when turning off, and the signal tends to go high for no apparent reason in the middle of a cavity-off state. If the cavity power were really fluctuating that much, we would expect it to be correlated with a change in the rates, but that doesn't seem to happen.

Fig. 3: Laser-state identification strip chart from Run 60732, including both trigger rates and cavity power variables.

It appears that we can't rely on the real-time variable for making this determination, at least during times when the cavity is empty. To identify the cavity on and off states, we will need to cut on trigger rate instead of on the real-time cavity power bit. We are concerned that the RT cavity power bit problem could potentially affect the old DAQ and its analysis as well, depending on the way it is wired and programmed. (Due to the passage of time and the fact that the original developers have left, this is something of a black box at present.)

Cavity Polarization Calibration

The laser polarization state is identified in the CMU DAQ's analysis software by requiring agreement between the EPICS and real-time variables (once, of course, their different conventions have been sorted out). To confirm that these variables did not suffer from the same difficulties as the RT cavity power bit, we prepared a strip chart of the two together (Fig. 4). (We have scaled the EPICS variable value by -0.5 and shifted it up by 0.45 in order to display the two together and remove the difference in conventions.) We can see that they agree well, except for the small (and expected) lag in the updates to the EPICS variable. At present, we are fairly confident in our ability to differentiate between the two laser polarization states.

Fig. 4: Stripchart of laser polarization variables from Run 60732.