We took 3 data sets with  fixed APV timing latency measured in 1/4 of RHIC clock ticks

3 attachemnts show all found pulses passing initial criteria ('pulses'), distribution of shap fit params vs. APV chip ('apv'), and selecton of valid pulses per FGT quadrant ('quad') 

Selection criteria for a puls to be processed (all other were simply ignored)

• status of the strip had to be 0 (good pedestal)  and it didn't get caught by oscillation and saturation filters. 
• at least one ADC > ped+600 
• fitted shape f(t,t0,tau,ampl) = ampl*(t-t0)^2*exp(-(t-t0)/tau)  , see last two attachments for visualisation

Fig 1.  Typical time shapes found by Len, 1 time bin has ~27 ns length

Fig 2. Accumulated raw spectra for FGT disc 1, for 'best' timing of 64 , run 13036072

Added:

Pulse shapes for formula with varied power of t  and fixed tau and t0:

f(t,t0,tau,ampl) = a*(t-t0)^p*exp(-(t-t0)/tau)  (+b)