How to produce a proton histogram

1

• How to produce a proton histogram
• By Erik Andersson Sundén Henrik Sjöstrand

2
Outline

• Introduction
• Glitches
• Baseline Reduction
• Gates
• Background Subtraction
• Gain Shift Corrections
• Proton Histogram

3
Introduction

• Signals are originating from two PM-tubes
  connected to the phoswich scintillators of MPRu.
• They are collected by the transient recorder
  card in the cubicle. (ADC with 0.5 ns time step,
  each volt bin is 4mV)
• This presentation is intended to clarify what we
  can do with our data since every event is
  recorded individually.

LED event
Proton event
4
Glitches

• Glitches have been observed on the baseline of
  channel 26 and 31.
• The amount of glitches increases with temperature
  in the cubicle.
• We can compensate for this on the baseline but
  not on the signal.
• More cooling in the cubicle is needed!

A typical glitch is shown with the LED event
above. The histogram below depicts the
difference of neighbouring bins on the baseline
before the event.
5
Baseline Reduction

• Where the concepts of the evaluation tool of our
  baseline reduction is presented.

6
Baseline Reduction 1
Mean
Line

• The signals from the PM-tubes have a pickup.
• Each channel has been analysed individually to
  find the best approach to it.
• Four methods were evaluated mean, line,
  polynomial and sinusoidal baseline reduction
  method.

Polynomial
Sinusoidal
7
Baseline Reduction 2

• Each channel is examined individually using the
  intervals around the event interval.
• A quality value, Q, is calculated of the middle
  interval. (n in the equation to the left is the
  number of bins in the middle interval.)

8
Baseline Reduction 3

• The Q value is calculated for each combination of
  lengths of the three intervals.
• To find the best method, given the length of the
  intervals, we choose the method giving the
  smallest standard deviation of Q.
• We can find what intervals that give the smallest
  standard deviation and which method to use.

A typical Q-value plot with different middle
intervals on the x-axis and fixed interval
lengths before and after.
Channel 30 for the polynomial baseline reduction
method.
9
Baseline Reduction 4

• The signal of any event is very long.
• This makes it hard to find a solution since we
  want the interval after the event to be signal
  free.
• Have to choose
• Long interval to estimate
• or
• Only use an interval before the event.

10
Baseline Reduction 5
Summary of Baseline Reduction

• Every channel has been evaluated individually
• If the background situation of a channel changes
  drastically this evaluation needs to be redone.
• Each channel has its specifically set baseline
  reduction method.
• We use an interval before the event to estimate
  the baseline under it.

11
Gates

• Where we will see that adaptive gates is a useful
  tool in our data reduction.

12
Gates 1

• Gates are used to identify events
• The short gate, Qshort, is now set to length 6.
• The long gate, Qlong, is now set to length 13.
• How do we find the best lengths?

The (fixed) gate intervals in this picture are
set as Qshort (46,51) and Qlong (52,64)
13
Gates 2

• When we use these gates a Qlong, Qshort-plot ca
  be produced.
• The Proton is quite well separated to the left
  and the LED is clearly separated.

14
Gates 3

• But
• The trigger is not in a fixed time bin. It seems
  to vary between two bins.
• To solve this we introduce adaptive gates
• maxmax of event
• Qshort(max-1,max4)
• Qlong (max5,max17)

The cards are not trigging in the same time bin.
The trigger level is 31 on this channel and the
events are expected to trigger in bin 46.
15
Gates 4

• The introduction of adaptive gates squeezes the
  proton island to be more narrow.

Adaptive Gates
Fixed Gates
16
Gates 5
Summary of Gates

• The gate technique is working fine to identify
  the proton island.
• We need to evaluate which gates to use.
• Adaptive gates seems to improve the
  distinctiveness of our proton island.

17
Background Subtraction

• Where we will see how the background is
  subtracted from our signals and how we finally
  end up in a proton histogram.

18
Background Subtraction 1

• Background data set is collected on JET pulses
  69472-69719.
• We are using the 14 MeV setting of the magnet as
  a background.
• In this way we simultaneously collect 14 MeV data!

14 MeV neutrons
Shown is the background data set of JET pulses
68421 68484.
19
Background Subtraction 2

• We can analyse the data by using a Qtot plot. The
  Qtot plots shown in this presentation is using a
  diagonal cut in the Qlong,Qshort plot including
  the events where Qshortlt2Qlong.
• Qtot Qshort Qlong

20
Background Subtraction 3

• Wanted to use the KN1 data to normalise the
  background to the signal data. Not successful!
• We are now scaling the background to the signal
  in a chosen region (blue).
• The proton region (green) is found by eye.

Qtot plot with the signal data (black) and
background data (red).
21
Background Subtraction 4

• Subtracting the (normalised) background from the
  signal data and counting the number of events
  finally gives us the number of proton counts in
  the specific channel.

22
Background Subtraction 5
Summary of Background Subtraction

• The background subtraction is working well.
• The current routine uses predefined intervals in
  Qtot to normalise the background to the data and
  to find the proton peak.
• An examination on why KN1 is not scaling with
  our data is needed.

23
Gain correction

• The position of the proton peak is dependent on
  the system gain (G)
• The gain has to me monitored and corrected for

24
The control an monitoring (CM) system

• The CM-system consist of a YAP source in channel
  0 and a LED signals in all channels. If the LED
  is not stable over time the LED light output
  (LED_lo) has to be monitored.
• Knowing the LED_lo, the gain for the different
  channels can be calculated

Channel 0
Channel x
25
Uncertainties in the CM-system
There is a statistical uncertainty in determining
the different quantities (assuming Gaussian
distribution)
26
The LED stability (1)

• The LED is not stable over longer time periods!
  It seem to correlate with the LED temperature.

27
The LED stability (2)
However normally no pulse to pulse instability
has been observed. The measured variation (0.42)
is almost the same as the one expected from the
statistical uncertainty (0.36).
28
Gain stability result (1)
29
Gain stability result (2)

• C2 the gain change from pulse to pulse. C3
  ?LEDmean.
• C4 ?gain.
• C5 the gain change between 68100 to 70353.

30
Gain stability result (3)

• The decrease in gain could be due to a decreasing
  amplification (in the PSA or in the PM tubes) or
  due to decreasing number of photoelectrons. If
  the number of photoelectrons decreased the
  relative width of the LED peak would increase.
  This has not been observed.

CH 9 (typical)
CH 23 (not typical)
31
Implementation

• The change in gain has been implemented into the
  data analysis. The raw data (pulse shapes) is
  divided with the gain value to compensate for the
  change in gain.

68000
70000
70000 With gain
32
Proton histogram
The Proton Histogram
Channel number on x-axis and counts/mm on y-axis.