We can see here that Rob's combination of sample rate and sample length is not optimal, because he is not capturing the lowest and highest point of the pulse.
- Cropped pulse
1) A frame of say 11 audio samples is read from the audio stream
2) Is the 6th. sample the peak value and also above the threshold?
example: ....2,3,1,5,2,3,[0,0,1,2,3,4,3,2,1,1,1],1,2,3,2,4,2,...
3) If NO move the frame one sample to the right, if YES capture the pulse
4) Normalise the pulse - this means factoring the samples such that [s + s + s + s +s ....] = 0
5) Calculate pulse distortion and compare it to the mean
6) If distortion exceeds the tolerance, simply drop the pulse otherwise;
5) Take the biggest sample value minus the lowest sample value to get pulse height in arbitrary units.
- Pulse height calculation
So to achieve good pulse height resolution you want to make sure you consistently catch the high and low points.
Think of the high tension wire going to the detector in terms of a spring and the energy of the PMT pulse as deflecting the spring.
We are mainly interested in how much the spring deflects.
Impulse gives you several parameters to achieve optimum results.
1) Sample rate - faster sample rate provides a smoother curve
2) Sample size - typically proportional to sample rate, make sure you see the critical part of the pulse.
3) Pulses to sample - How many samples to catch before calculating the mean pulse shape
4) Buffer size - use a larger buffer size for higher sample rates.
5) Peak shifter - shifting the peak to the left or the right can make it easier to get both the peak and the trough into the frame.
Finally we want to know how much distortion we have in our system, so we run the distortion curve on a number of samples.
This function gets a bunch of pulses as determined by "Pulses to Sample" and orders the result in order of distortion, lowest to highest.
Hope this explanation is clear and understandable.
Steven
