A Closer Look at Beam Profiles (Modified)

Using the most recent run of 100 sequential readouts, I have plotted a few of the beam profiles. The overlaid profiles fit very closely with one another, so I also created a histogram which gives the mean value of each bin based on the five profiles used, then plotted deviations of the profiles from the mean.

Reliability and CCD Correlation

I ran a new test using a script to readout from the DCOPS 100 times. Since I didn't have to be directly at the table, I eliminated the possibility of bumping the equipment and skewing the results in that way. I also looked at correlations between the readings from the parallel pairs of CCDs. The horizontal CCDs (1 and 3) show a standard deviation of less than 1.2 pixels and little correlation in the deviations. The vertical CCDs showed a larger standard deviation of 2.0 pixels, but greater correlation. Some aspect of the setup must be causing extra vertical fluctuations.

Results are presented here.

Peak Fitting and Translation

I have completed two new studies of the reliability of the peak fitting algorithm across the length of a CCD. The results show a linear relationship between sensor displacement and fitted peak center up to the final 100 pixels of the 2048 pixel CCDs. This corresponds to about 1.4 mm out of a length of 30 mm. I can't draw any clear conclusion about correlations between deviations in the two CCDs.

Graphs and summaries are given here:
Test 5
Test 7

Translation Stage Calibration

The damaged readout board has been fixed by Valery at CERN, and I am now able to get readout from our local DCOPS again. In preparation for more demanding tests using the translation stage, I've run a calibration so that we can measure distances by number of turns of the translation knob rather than by reading from the attached ruler.

As a pessimistic estimate, readings from the attached ruler are good to 0.5 mm. By marking the knob on the translation stage, I get a resolution of about 1/80 turn. For the calibration, I ran the stage through its entire range, counting the number of turns and taking ruler readings at the first and last turns. I performed one trial in the positive direction, and one in the negative. The results are as follows:
Trial 1: Start at 28.58 cm, End at 15.48 cm, 67 turns
Trial 2: Start at 15.45 cm, End at 28.52 cm, 67 turns

Propagating errors, trial 1 gives (1.955 ± 0.011 mm/turn) and trial 2 gives (1.951 ± 0.011) mm/turn. Averaging gives a general result of (1.953 ± 0.008) mm/turn for a translation over the entire stage. In the translation tests, then, if we always move the stage in increments of one turn, the 1/80 turn uncertainty dominates, leading to an uncertainty in position of 30 microns. Thus, we have achieved a resolution on the order of a few CCD pixels (one pixel being 14 microns wide).

Electronics Repair

Progress on reliability tests has been halted due to damage to the interface board when I was trying to compensate for extra current draw with all three readout boards attached. In the past week, I have been working on diagnosing the issues and trying fixes, but the eventual solution was to ship the board to CERN for Valery Andreev to look at. There appears to be a short somewhere in the circuitry, lowering the resistance to 25 ohms so that it draws nearly 250 mA of current.

Background Reliability

Graphs given here. In this test, a new background was taken between each trial. For amplitudes, RMS ranges from 1.8 to 3.6; for peak locations, RMS ranges from 1.7 to 2.1. These are in general slightly wider than the two previous tests, but again the sample size is too small to identify any significant difference.

More Reliability Results

Histograms from another test using our newly acquired laser can be seen here. This time, the laser was left off for approximately fifteen seconds between trials. The resulting distributions are comparable, or sometimes tighter, than those when we used a constant beam throughout. The RMS for the amplitude distributions here are between 1.1 and 2.2 as opposed to 2.0 and 4.0 previously. The RMS for the peak location distributions here are between 1.3 and 2.1 as opposed to 1.2 and 1.4 previously (note that the 2.1 value, while bigger, comes from a CCD 2, which wasn't recorded previously).

The effect of cycling power to the laser seems negligible, at least on this small of a time scale. If anything, it looks as though the beam may drift more when it stays on continuously, but the amount of data is not sufficient to draw that conclusion.