1. HELIX time of flight system electronics concepts
Gerard Visser
Indiana University CEEM
HELIX design Univ. of Michigan, Feb. 12th, 2016

2. Dark count rate should be as low as possible
General considerations
For optimum timing we must pay careful attention to (at least) five issues
Scintillator and optics: deliver as many photons to sensors as possible, with minimal skew
Actually though only what happens within the response risetime of a single- photon pulse at discriminator matters – quantity and time skew of later photons is irrelevant
Electronics noise in terms of photons, divided by slew rate of a single-photon pulse at discriminator, should be as low as possible
Note that SiPM’s enormous capacitance makes this difficult in several ways, compared w/ PMT’s
Note also that fast is good, but there is such a thing as too fast: The electronics noise generally rises faster than linearly with bandwidth, the slew rate does not (and certainly does not when you hit physical limitations of the sensor). Again this is rather different than with PMT’s.
Dark count rate should be as low as possible
It contributes jitter just as electronics noise does
Perhaps more importantly, higher dark count rate may force higher timing discriminator threshold to avoid pileup problems with the TDC
Skew in electrical connections, and random delay (“TTS”) of sensors should be as low as possible
If electronics noise and/or dark count noise is dominant, you do not gain by adding sensor area. Number of detected photons, dark counts, and electronics noise all rise in proportion.

3. A large area SiPM is a very low impedance detector 
Collecting the charge
generic SiPM model
 distributed inductance & resistance not shown 
fired pixels (few, proportional to light signal)
spectator pixels (many, proportional to active area)
CBYPASS
A large area SiPM is a very low impedance detector 
For TOF, we must collect the signal charge as fast as possible – or at least as fast as the optimal risetime to discriminator.
Preamp impedance and bypass capacitor impedance must be low compared to spectator pixels! E.g. 6 5 ns risetime  0.4 Ω !!
Series connection can help…
Load impedance model (preamp)

4. All SiPM mounted close packed One large light guide
Bar, light guide, and SiPM mounting scheme
Bar 10 cm × 1.5 cm
6 × SiPM (0.6 cm)2 each
All SiPM mounted close packed
One large light guide
SiPM 1.67 cm pitch
One small light guide per SiPM
Might be mechanically coupled
Expect 6 × shorter
half as wide?
Optical simulations should be used to decide which is best…
Lowest time skew (for first photons – not all!)
Light collection efficiency (again, for first photons)
I think 2nd option is better for the SiPM board assembly, and I am guessing is maybe best performance too. B A
FEE board location options (I prefer option A) C
SiPM board, glued or clamped to bar & light guide
FEE board estimate 9.5 cm × 12 cm

5. One small light guide per SiPM
For comparison – STAR Forward Preshower Detector
244 bars, one end read out
Bars 4 cm × 1 cm, 5.8 cm × 1 cm
2 × SiPM (0.3 cm)2 each
(for electron, photon identification – no TOF)
Length ~1 m (various)
SiPM 2.0, 2.9 cm pitch
One small light guide per SiPM
FEE board 3.8 cm × 6.7 cm
Note: Light guide exit slightly smaller than SiPM active area
FEE board
FEE board with temperature compensated bias regulator, current monitor, preamp and cable driver
SiPM board

6. DCR × crosstalk 1-pix Q / capacitance Manufacturer SensL Hamamatsu
SiPM candidates
µFJ TSV
S VE ??
Manufacturer
SensL
Hamamatsu
Active area
(6.07 mm)2
(6.00 mm)2
PDE nm nm
Dark rate (typ.)
1.7 MHz (2.0 MHz ??)
2.0 MHz
Crosstalk
7 %
3 %
“Gain” (1-pix Q)
2.8 × 106
1.7 × 106
@ overvoltage
2.5 V
3.0 V
Capacitance
4.0 nF ??
1.3 nF
Recharge τ
48 ns
~30 ns ??
Package
TSV CSP BGA
TSV LGA
Price / bar end
$ 348 ?
2 pix DCR
119 kHz
60 kHz
Cap. FOM
112 µV
210 µV
for reference only; there is no particular requirement
DCR × crosstalk
1-pix Q / capacitance
derived params

7. TIMING WAVEFORM GENERATOR
HELIX TOF frontend board – one per bar end
very low input impedance
(R values here from nEXO version – ignore!)
pink boxes are 6× , i.e. 1 per SiPM chip
I MON ADC
SLOW SHAPER
ADC
VOLTAGE REGULATOR
SiPM
COMMON BASE PREAMP
TIMING DISCRIM.
TIMING WAVEFORM GENERATOR
ADC
on SiPM
board
DAC
DAC
THERMISTOR
TRIGGER DISCRIM.
TRIGGER LOGIC
DAC
DAC

8. HELIX TOF FEE BOARD – OPTIONS TO CONSIDER
Timing measurement
HELIX TOF FEE BOARD – OPTIONS TO CONSIDER
(I do not necessarily advocate)
Instead of using ADC board for timing, can use some other TDC
Slow shaper signal can be sampled/digitized on this FEE
either in response to local trigger (at <200 kHz, say)
or in response to global trigger
Default plan for TDC by ADC is a ringing waveform few cycles of about 33 MHz (for 100 MSPS ADC), then “Q-switched” to damp quickly to zero. Overall width should be <500 ns.
 can get timing easily on pulses that do not overlap… (Here state slat-end 2-pixel dark rate and then overlap 500ns)
what else to say?
2π × 33 MHz × 90% × 1024 = 0.19 ADC units / ps
2π × 27 MHz × 90% × 4096 = 0.62 ADC units / ps
2π × 22 MHz × 90% × = 2.0 ADC units / ps
Can do a nice <10ps TDC with this, if we can generate a waveform and transport to the ADC chip with <10ps jitter. Yes we can!

9. Occupancy from dark counts
Each channel (slat end) has 216 mm^2 of SiPM area
Assume Hamamatsu SiPM’s
1-pixel dark count rate 12 MHz, 2-pixel DCR 360 kHz, 3-pixel DCR 11 kHz
Readout window in TOF can be very narrow, but for data planning purposes let’s assume not so narrow: 500 ns
 occupancy 1-pixel, 2-pixel, 3-pixel
Conclusion
We should plan on ZS at 3-pixel signal level
Maybe this is the proper threshold for HELIX TOF trigger too?
Dark count occupancy is negligible
Real hit occupancy = 3 + occasional few more, is that right??