The results from the timing scan performed on 02/06/12

Run numbers are 130370??, where the last two digits are: {25, 26, 28, 29, 30, 33, 34, 35, 36, 37, 38, 39, 40}.

The corresponding latency used to change the time offset: {30, 33, 39, 42, 45, 48, 51, 54, 57, 60, 63, 66, 69}.

 Observed pulses were fitted with this function 
f(t,t0,tau,ampl) = ampl*(t-t0)^2*exp(-(t-t0)/tau)

There are two addition plots added to the apv_* files, showing the distributions of the minimum and maximum time bins. 

For all plots, adcmax > 600+ped, ped = adcmin. For some,  chi2 < 2.

Since the gross timing behavior is well defined for all apvs except for couple of odd ones, I integrated the t0 distribution over all apvs, and plotted it against the run index, which is essentially increasing latency value with almost uniform steps. (one step is missing, I think)

Fig. 1. Gross t0 distribution vs. run index normalized by events per run

So it seems that we are scanning through multiple crossings, and the step size here is quite big. The "bimodal" behavior we saw previously is now almost certainly seeing the tail of the previous collision. Since the time bin size is ~27ns, and the step size (I think) is 3 bins, it makes sense that after two steps (6 bins ~ 162ns ~ 1.6 crossing) we'll see the next crossing.

Fig. 2. Fraction of events with t0 between 1.5 and 3.5, the indddex is binAPV but I integrated over every 10 apvs, making it essentially vs. quadrant.

At least for me, it's hard to make sense of this plot, because the step size is quite big and we do not get a smooth behavior. I would pick latency=48 and fine tune around that point?

Fig 3, comparison of t0 for all latencies for selected assembly. A repetitive pattern every 2 RHIC clocks suggests FGT sees mostly beam halo from 56 bunches in the ring.  

Fig. 4. Same as figure 1, but only using R strips from 20~120, and 420~520.