Signal Shape and Signal Loss in MaPMTs

1
Signal Shape and Signal Lossin MaPMTs

• Fit of MaPMT signal spectra
• 1st dynode effect in CAMAC data
• loss definitions
• gain variations
• Signal loss
• HV scan data ranges and control plots
• resistor chain options with different gain
• readout options with different attenuation
• loss behaviour
• Conclusions

LHCb RICH meeting CERN, 28.5.2002
Stephan Eisenhardt University of Edinburgh
2
Fit of MaPMT Signal Spectra

• Poisson fit up to 2 photoelectrons from photo
  cathode (PC) or 1st dynode (1st) (sum over
  Gaussians with Poisson weight)
• Gaussian fit for 0 (pedestal) or 3 photoelectrons
  (PC)
• Signal width (?1) hardly constrained
• Bug fix in normalization improved fits
• Problem with APVM data
• jitter of LED light source large
• some signals only partially sampled
• npe(1st)/npe(PC) gt 0.5 too high!!
• jitter looks similar to 1st dynode effect
• the 1st dynode effect neither can be confirmed
  nor excluded from current APVm data

APVM data
typical run from HV scan series
3
1st dynode effect in CAMAC data

• Poisson fit better overall fit but region
  between pedestal and signal still not described
• Poisson 1st dynode
• reasonable description of data
• ? (1st)/?(PC) 0.10.3 believable
• gain at 1st dynode (K1) matches expectation

• for CAMAC data
• gate width 200 ns
• signal fully contained
• still events between pedestal and signal
• but no distinct peak (even at highest gain)
• Gauss fit bad description

Poisson fit
Poisson fit 1st dynode
Gauss fit
? 0.24
?PC 0.215
? 0.24
?1st 0.025
g1 ? Q12/?12 3.2
K1 4.1
K1 6.1
CAMAC data HV scan run 2041 HV -900V
4
CAMAC data 64 pixel
Gain at 1st dynode calculated for Gauss add 2.5
no multiplication at 1st dynode fitted for
Poisson Loss consistent definition based on the
photo- conversion at PC (?) relative
change consistent with change of fit method
Gauss fit
Gauss fit
ltg1gt ? Q12/?12 1.9
ltlossgt 0.105
Poisson fit
Poisson fit
ltK1gt 2.8
ltlossgt 0.156
Poisson fit 1st dynode
Poisson fit 1st dynode
ltK1gt 6.0
ltlossgt 0.132
gain at 1st dynode
signal loss below 5?-cut
CAMAC data 64 pixel of one tube focussed light
source HV -900V
5
Loss Definitions

• Signal loss can be defined as
• loss of photoelectrons converted at the
  photocathode
• loss of photoelectrons converted at either the
  photocathode or the 1st dynode
• An estimate can be derived from fitted
  parameters
• photon conversion probability (?Poisson)
• (? is a bad approximation for 1-e-?, to be fixed)
• integral over the 1-photon contribution(s)
• (used in this talk from
  now on)
• integral over the pedestal contribution

APVM data
among the best signal resolution from HV scan
series
6
APVm Data Sets and Selection

• Available data with APVM read-out
• HV scans
• resistor chain ratios
• standard gain 3-2-2-1-.-1-5
• medium gain 4-2-2-1-.-1-5
• low gain 4-3-3-1-.-1-5
• (1/50) or no attenuation before APVm
• old and new focussing in MaPMT
• center and border pixel of MaPMT
• Simultaneous illumination of 64 pixels
• taken in magnetic field setup
• old and new focussing in MaPMT
• HV -900V, -1000V
• All further shown data was taken with APVm
  read-out

• Every fit (600) inspected and judged for
• general fit
• ill behaviour of components
• HV scans
• HV range determined in which the fits are
  regarded as reliable
• further restriction of range if runs behave bad
  in control plots
• 64-pixel illumination
• 18 runs refitted with slightly different start
  parameters to improve general fit

7
Gain Variation

• within tube
• gain variation (max/min) lt 4 (spec lt5)
• but the two halves behave differently and in the
  same way for different tubes
• caused by different capacities of two Kapton
  cables
• gain variation most probably 2
• between tubes
• gain variation lt2

gain
old focussing
new focussing
APVM data magnetic field setup
8
Control Plots I (std. att.)
Wide dynamic range S/N gt70.4 good fit
behaviour K1 6 (scaling washed out) Loss ?
may become negative for PC only (not all
contributions regarded) overestimated due
to approx. of 1-e-? 1-phe positive definit
well behaved upper limit (only higher
multiplicities neglected) ped regards all
signal contributions
old focussing
fit parameters
derived parameters
9
edgy fits for no att. option
upper edge of dynamic range
lower edge of dynamic range
signal starts to vanish in pedestal

• non-linear (strange) behaviour of FED
• depopulation of higher ADC channels
• with increasing gain
• mirroring into lower dynamic range

10
Control Plots II (no att.)
Dynamic range limited S/N 30.10 relatively
bad fits K1 6.3 (scales) Loss seems to be
generally higher but is due to worse S/N
new focussing
fit parameters
derived parameters
11
Resistor Chain Options - Gain
standard attenuation
no attenuation
fits become edgy
PC
1st

• very limited dynamic range
• maximum gain lower
• limited S/N

gain ratio std./low gain 4
12
Resistor Chain Options S/N
standard attenuation
no attenuation
S/N gain
PC
1st
lower gain seems to have better S/N (pedestal in
the data is narrower)
high gain provides better S/N
13
Resistor Chain Options Loss(1-phe)
standard attenuation
no attenuation
PC 1st
PC
good match between options loss scales with S/N
low gain option has lower loss due to better
S/N (pedestal in the data is narrower) loss
figures generally match the std. att. case
14
Resistor Chain Options Loss (std. att.)
standard gain
low gain
PC 1st
PC
no difference between tubes (old vs. new
focussing) std. gain option clearly preferred
(better S/N ? less loss)
15
Resistor Chain Options Loss (no att.)
standard gain
low gain
PC 1st
PC
no difference between tubes (old vs. new
focussing) low gain option gives better S/N ?
less loss
16
Conclusions

• Description of signal spectra 1st dynode effect
  needed for reasonable fit
• in APVm data light source jitter masks 1st
  dynode effect
• in CAMAC data no clear peak either but signal
  contained in 200ns gate
• and there is data between pedestal and signal
  peak!
• Loss behaviour
• scales with S/N
• good match between resistor chain (gain) options
• with attenuation standard gain preferred (higher
  S/N ? lower loss)
• without attenuation low gain has better S/N
  (better pedestal) ? lower loss
• No attenuation option not preferred
• lower gain ? lower S/N ? higher loss
• very limited dynamic range
• To be done
• improve LED light source
• verify 1st dynode effect with APVm read-out
• add error calculation

17
(No Transcript)
18
24.1.2001 Signal Loss

• LED Pixelscan, CAMAC
• ltSignalgt 69
• ltNoisegt 1.98
• ltpedgt 61
• ltS/Ngt 37
• Signal Loss
• f (N-n)/n
• N trig (1 - exp (-?))
• n trig gt ped 5 ?
• Result
• ltf gt 10.2 (3.9 RMS)

Signal Loss vs Signal/Noise
Signal Loss
Signal/Noise
19
24.1.2001 Threshold Spread
APVm frame
CAMAC
Noise

• Principle treat data as if they were binary
• Threshold spread
• RMS 5 .. 8 ADC
• Set binary threshold
• thr ltpedgt 4 RMS 20 ADC counts

Pedestal
20
24.1.2001 Binary Signal Loss

• Measure signal loss with constant threshold
• removed few channels, large noise or threshold
• Signal loss results
• constant threshold ltf gt 19.5 (6.0 RMS)
• 5 ? cut ltf gt 10.2 (3.9 RMS)
• Binary signal loss
• smaller than 20
• Worst case assumption
• Beetle 4 bits for threshold

Signal Loss
5 ? cut
Binary
21
24.1.2001 Signal Loss Simulations

• Monte Carlo simulation
• preliminary
• based on data, e.g. pixel-to-pixel gain
  variations
• gain g dVk
• Poisson at 1st 3 dynodes else Gaussian
• does not (yet) reproduce small signals
• Signal loss results
• constant threshold ltf gt 9.2
• 5 ? cut ltf gt 5.0 )

Data vs Simulation
Data
Monte Carlo