It was found that there were unknown peaks in ADC spectra of all sectors of the Downstream Collar. These events were presented in the different triggers: Kp22, Km21, piscat. To understand an origin of these peaks a special study was carried out for Kp22 triggers. First of all, the usual kinematical parameters, such as Ptot, Rtot, tk and tpi confirmed a presense of decayed kaons. That might be meaning the peaks in Collar could  be caused not by the kaon decay products which have been detected by the RS and other subdetectors. The main sources of such peaks could be either the charged particles (e.g., MIP pions) from the beam or photons from Kp2 pi0 decays.

If this is true, then these particles should hit the EndCap crystals. Indeed, certain correlations between the hits in DN-Collar-2(elempvco=14) and in the EndCap module #3 (the inner ring) were found (as well as similar correlations for DN-CO-8 and EndCap-1stRing-module #9).

Then correlations of these selected events with hits in the I- and V-counters also were revealed (see vic.ps and vic1.ps ).
The first file (vic.ps) shows what are the rates of the I- and V-counters when a) there is no cuts (top left); b) there are hits in the EndCap module #3 of the inner ring (top right); c) events  with hits in the EndCap-3 and DN-Collar-2 (bottom left); d) events from "c)" with collar energy between 10 and 23 MeV (bottom right). The vertical axis shows a number of events per a I/V module; the horizontal axis corresponds to a I-counter module when the number is between (1-6) and to a V-counter module when the number is between (7-12). It can be seen from the bottom plots that there is a strong correlation  with hits in I-/V-counter #1 and #7 which correspond to I-counter module #1 and V-counter module #1 (1=7-6). Similar plots for the DN-Collar-8 are shown in the file vic1.ps

After looking at the Cherenkov-pion ring time the nature of the ADC peaks in Collar (Downstream) became clear. These peaks are caused by the pions from the beam, which hit the Cherenkov-pion ring, I/V-counter, Endcap(Down) and Collar (Down). This can be seen from the plot
http://www.phy.bnl.gov/~kmarat/collar/cpi_co.ps
or here (jpeg which quality is worse)

First histogram (top left) shows the ADC spectrum of the Collar(Down) module #2 (elempvco=14) when there are hits in the I- and V-counter and in one of the Endcap modules (Down). Statistics is not high, but the peak is clear.

The second histogram (right top) shows the Collar timing for the events in the peak region (from 10 to 23 MeV) of the first histogram. 86 events survived out of 152, and their distribution is flat in time.

Third plot (left bottom) is a scatter plot for 86 survived events: it shows Collar time versus time of the Cherenkov pion ring. The correlation is clear for the most of the events.

The last plot (right bottom) is for the Cherenkov kaon time for the same events. No correlation can be seen (or, better to say, there are not so many correlated events).

You can see similar plots for other modules:
Collar-DN-8 module  http://www.phy.bnl.gov/~kmarat/collar/cpi_co1.ps
Collar-DN-1 module http://www.phy.bnl.gov/~kmarat/collar/cpi_dn1.ps
=================
So, my conclusion is that, in principle, we can calibrate Collar detector (at least Downstream) by the beam particles which should deposit something around 25 MeV per a sector (by the way, I also found peaks in the Endcap individual module energy spectra). The statistics of this calibration could be increased by adding the data from different triggers with kaons (Kp22, Km22 ).
I understand, that we don't need too accurate collar energy calibration for the pnn events, but we can use this procedure to check the light yield of the collar detector.
==================
 

This is an example of such a test: a comparison of peak positions
(in ADC counts) of the Collar Downstream sectors for two groups of runs:
47562-48945 (beginning of 2002) and 49413-50213 (end of 2002 cycle).
 http://www.phy.bnl.gov/~kmarat/collar/comp1.ps

The horizontal axis is a DN-Collar sector number (1-12), the vertical axis shows the peak position (ADC counts) for the corresponding sector. Here, the solid line is for the beginning of the 2002 cycle (47562-48945 runs), and the dashed line shows how the peaks been shifted at the end of the 2002 cycle (49413-50213 runs). The error bar lengths were taken from the rms of the fitted peaks (beginning and the end, see the histograms from pages 7-12). The fit was done using Gaussian function. To save statistics the events in Collar sector were accepted if their time (tpvco) was close to the Cherenkov-pion ring time (cpitrs), which usually meant  -5ns<(tpvco-cpitrs)<1 ns.

It should be noted that the strange shape of the sector-to-sector dependence was caused by the fact that the HV for a new Downstream Collar was not lined-up (old Donwstream settings were used).

The conclusion from the picture above is that there is no visible degradation of the new Downstream Collar light yield, at least for 3-6 months of use.
 
 

What about peaks and correlations in the Upstream Collar? As one can see from the mentioned plots (beginning and the end) there is no clear peaks for Upstream Collar. In principle, it's understandable if to consider the geometry and the trigger demands. A new Upstream Collar is about to replace the old one.

================

Timing calibration of the Downstream and Upstream Collars was carried out by using pnn1or2 data (Ntuples). Example of the files which are needed for this calibration could be find here (based on Milind's scripts): pv_my.kumac and my.f
The resulting histograms fitted by flat background + gaussian are here: t0_dn.ps.gz (DN)  and t0_up.ps.gz (UP).
================
July 28, 2002.
Marat Khabibullin,
INR RAS