Extract a relative timing peak for each detector element from events with a corresponding hit in CPV element 28. To aid in overall event timing we'll extract the timing peak of hits in CPV-28 used above. These timing peaks have a run-dependence, specifically for six discrete sets of runs, which requires six map entries for each detector element. Those run sets begin (and will be put in the map at time value):
The wide TDC windows on the detector elements creates a large subtractive background in the timing offset plots, making it very difficult to extract tagger information. To reduce this background I put a 50-ns-wide window around the CPV-28 peak and cut hits falling outside. This cut was used for the BSD, BGV, and Tagger to maintain consistency.
9-26-02: The full data set has been processed through RODD v1.1, and the values are being extracted using a set of macros I am developing. The macro "pass_zero" currently runs well, but has some bugs while fitting. Recent improvements in my fitting parameters have reduced the problem, and I have first-guess values for the BSD and BGV constants, but there are several problem channels. I may need to study these specifically and make a second pass through the data set with fine-tuned cuts. Results are written to a text file labelled "pass0_[Detector]_[MapTimeEntry].txt" as a vector of F10.2 values.
BSD-S8 seems to be differently shaped from it's predecessor, causing the fitting parameters to deviate: jpeg image
BSD-R2 has a very heavy misshapen background in one runset: jpeg image
Double peaks are seen in BSD-R0 through R2 from runsets 2 through 6: jpeg image runset 2; jpeg image runset 4
BGV-8 has a stepped background in runset 3, not seen in other elements: jpeg image
10-04-02: I've gotten the fitter to run through the tagger channels for all six runsets, and have the results in .txt files containing vectors of timing offsets. If you want them, you can download them:
10-07-02: I've gotten the fitter to run through the taggerBGV channels for all six runsets, and have the results in .txt files containing vectors of timing offsets. If you want them, you can download them, as well as the CPV28 reference peaks I made by hand:
10-23-02: There seems to be a problem on grendl, putting histograms together. Below are plots of BSD R0 for the six runsets, the first taken from a ten-run segment during runset 5 and the second is supposed to be for all runs. The first should be a subset of the second, but it can be seen that the data from the first is not present in the second. I've re-compiled and re-tested the executable, and don't believe that the problem is within RODD.
10_23_02_01.jpeg
10_23_02_02.jpeg
11-27-02: I haven't come to any thrilling conclusions explaining the few problem channels I found in my analysis of the Tagger, BSD, and BGV:
3-08-03: The latest pass zero scheme is to extract timing offsets for the upstream detectors based subtractively on Tagger hits, rather than on CPV hits. The prescription set down by Richard can be found at: http://zeus.phys.uconn.edu/radphi/calibration/detector_params.shtml
The BGV results were obtained without much difficulty, and are available for six map entries:
Obtaining results for the CPV and BSD was more problematic. In the CPV, one sees a "jitter" in the signal that increases in amplitude as you approach high-rate central channels. The BSD displays multiple-peak structure, making extraction of a dependable timing offset difficult.
This algorithm uses tagger hits from all channels, offset from CPV 28. I thought that there might be a problem with this method, but the only effect I could see from limiting ourselves to one tagger channel was a loss in statistics. From there, I studied effects of closing the window on Tagger hits (currently set at +/- 40 TDC channels).
Closing the window does make the peaks easier to distinguish, but too tight a cut takes away real timing events. Other tools for getting rid of these problems had less helpful results. At this point I can obtain Pass Zero values for the CPV and BSD, but am not confident in the accuracy of any fit made to the timing spectra.
3-18-03: As per Richard's and David's suggestion, I'm doing my best to ignore the above effects and fit over them, rather than try to suppress them. We have no good explanation for the multi-peak structure in the BSD channels, but are confident that the jitter in the high-rate CPV channels is due to 2-ns beam structure. The timing and event rates of the Tagger and CPV are much better/higher than the width of the windows, allowing correlated hits to build up over time. Viewing a subtractive TDC spectra allows us to see a trend of 2-ns pulses.
I expect to have new timing values shortly....
