Transcript Document

Trip-t and TFB status
Trip-t
brief description of internal architecture and interfaces
proposed Trip-t operation at T2K
SiPM connection, gain and discriminator threshold considerations
Results from latest Tript version
linearity and discriminator measurements
TFB prototype status
results from prototyping elements
ADC functionality and test results
HVtrim functionality and test results
Calibration circuitt description and test results
TFB layout status
TFB firmware status
future plans
DRAFT TALK – NOT YET FINISHED
Mark Raymond - 27/6/06
1
Trip-t single channel front end architecture
very simplified – neglecting features not relevant to ND280 operation
preamp
integrate/reset
gain adjust
1,2,3,…8
analogue pipeline
Qin
1pF
3pF
gain 1 or 4
discriminator
x10
disc. O/P
Vth
reset
only preamp gain affects signal feeding discriminator – no fine control (x1 or x4)
Vth common to all channels on chip
analog bias settings, gain, Vth, programmable via serial interface
Mark Raymond - 27/6/06
2
Trip-t full chip
simplified and neglecting features not relevant to operation in ND280
48
top
16 IP/s
bottom
16 I/Ps
32 analog
outputs
32
front
end
chans
32
analogue memory
(pipeline)
control
32
32:1
analog
MUX
serial
analog
output
control
top 16 disc. O/Ps
bottom 16 disc.
O/Ps
bias, control, reset
dig.MUX
32:16
control
top or bottom
16 disc. O/Ps
serial programming
interface, bias gen.,
control interface, …
dig.control
32 channel chip -> 1 serial output, 48 deep analogue pipeline to store sampled front end outputs
(note: pipeline operated using 2 timeslices/preamp integration period, so length reduced to 23
see http://www.hep.ph.ic.ac.uk/~dmray/pdffiles/FIFOtalk_1_3_06 for detailed explanation)
have to select either top or bottom 16 disc. O/Ps to transmit off-chip
~ 12 digital control/programming inputs, 16 disc. outputs => ~ 30 I/O lines/chip (2.5 V CMOS)
Mark Raymond - 27/6/06
3
Trip-t operation at T2K
Proposed mode of Trip-t operation for beam spill data acquisition is as follows
during spill
integrate signal for each bunch and store result in pipeline* (15 timeslices for 15 bunches)
timestamp high gain channel discriminator outputs that fire
after spill
continue running in same way, for a while, to catch late signals (m decay)
readout entire contents of pipeline
assemble data block containing hit timestamps and all digitized analogue data and transmit
transmitting all info in this way allows histogramming of single p.e. events to monitor SiPM gain
vast majority of data is pedestal + single/double p.e. hits only
start of spill
5.25 ms
spill period
at this time trip-t switches
to inter-spill operational mode
(cosmic trigger)
end of spill
2.8 ms
after spill
active period
Mark Raymond - 27/6/06
74 ms (23 cell) readout period
(if O/P mux running at 10 MHz)
4
Tript for ND280, gain considerations
need ~ 500 p.e. dynamic range, while simultaneously discriminating signals at the ~ 1p.e. level
can’t be done with one gain range => split signal between high/low gain ranges (channels)
Signal shared between Cadd, Chi and Clo (also some strays), Chi/Clo = high/low gain ratio
HVglobal
simplified single SiPM channel schematic
Chi
100pF
trip-t
1 MW
Clo
10pF
thin coax
50W
SiPM
HVtrim
Cadd
330pF
Choose Cadd to match final SiPM gain (330pF about right for 5x105)
Cadd also helps with gain discontinuity when hi gain channel saturates
(see http://www.hep.ph.ic.ac.uk/~dmray/pdffiles/tript_talk_1_3_06)
don’t know what final SiPM gain will be, but assume production devices will be quite well matched
in any case will have individual channel gain adjustment by HVtrimDACs
Mark Raymond - 27/6/06
5
Discriminator threshold (Vth) considerations
1pF
analogue
pipeline
Qin
reset
disc. O/P
x10
Vth
Vth only relevant to the 16 high gain channels
- remember only 16 channels can be selected for transmission off-chip, so just arrange for
these to be the high gain channels
(Vth also applied to low gain channels, but we don’t need to look at the outputs of these)
Vth needs to be set high enough to prevent single p.e. events triggering discriminator
(otherwise single p.e. triggers will dominate and will lose ability to timestamp real signals)
uncertainty in threshold setting given by spread in discriminator turn-on curves across chip
can choose high gain channel value (external capacitor division ratio) but trade-off between
threshold adjustment range and uncertainty in threshold value
Mark Raymond - 27/6/06
6
Gain and gain ratio considerations (1)
1pF
single tript
channel
analogue
pipeline
Qin
reset
disc. O/P
x10
Vth
~ 1V dynamic range
available at preamp O/P
~ similar voltage range
at x10 amp O/P
~ similar disc. thresh.
voltage adjustment range
2.5 V CMOS so can assume dynamic ranges of internal circuits ~ 1V
this has implications for discriminator threshold range
if want 0 – 5 p.e. adjustment range then 5 p.e. ≡ 1V at x10 O/P
=> 1V ≡ 50 p.e. at preamp O/P
so high gain channel will saturate at ~ 50 p.e.
this translates to threshold uncertainty ~ +/- 0.5 p.e. (measured – see later)
Mark Raymond - 27/6/06
7
Gain and gain ratio considerations (2)
HV(TFB)
simplified single SiPM channel schematic
Chi
100pF
Trip-t
1 MW
thin coax
50W
Clo
10pF
SiPM
HVtrim
Cadd
330pF
So discriminator threshold range adjustment 0 -> 5 p.e.
High gain channel saturates at 50 p.e.
Choose Chi/Clo so low gain channel saturates at 500 p.e.
Note: These values are examples and can change, but need to take care with
threshold adjustment range/uncertainty trade-off
Mark Raymond - 27/6/06
8
Latest Trip-t test results from final version
2nd (final) tript version very similar to 1st
minor architecture change to improve
O/P stage linearity
12
Trip-t V1/V2 linearity comparison
version 2 linearity clearly better but
still some gain reduction for small
signals
3
10.5x10
11
 will need electronic calibration to
correct for linearity
3
x10
ADC units
10.0
10
9.5
version 1
version 2
9.0
9
8.5
0
Mark Raymond - 27/6/06
1
2
3
Qin [pC]
4
5
9
Tript V2 linearity(1)
all 16 channels, hi and lo gains
4000
higain
cahnnels
component values chosen for
SiPM gain ~ 5x105 (Chi = 100pF
Clo=10pF, Cadd=330pF)
3000
ADC units
lo gain saturates at ~ 40 pC (500 p.e.)
hi gain saturates at ~ 4 pC (50 p.e.)
logain
channels
2000
1000
0
0
10
20
30
40
total external charge injected [pC]
Mark Raymond - 27/6/06
10
Tript V2 linearity(2)
4
log-log plot of same data
3
2
10:1 gain ratio means gain range
change occurs where logain signal
size already large so no S/N problems
higain
channels
1000
8
7
6
5
ADC units
4
logain
channels
3
2
100
8
7
6
5
4
3
2
10
5 6 78
2
3
4
0.1
5 6 78
2
3
4
5 6 78
1
2
3
4
10
total external charge injected [pC]
Mark Raymond - 27/6/06
11
Tript V2 discriminator measurement
discriminator curves for all 16 higain channels
count the no. of times the discriminator fires
for 1000 preamp integration periods
sweep the injected signal size
# of times disc. fires
for 5x105 1p.e. -> 0.08 pC
1000
pk-pk width ~ 1 p.e. also for this measurment
so +/- 0.5 p.e. precision
can improve precision but remember trade-off
with adjustment range
800
600
400
200
0
0.00
0.05
0.10
0.15
0.20
0.25
0.30
0.35
total external charge injected [pC]
1 p.e.
Mark Raymond - 27/6/06
2 p.e.
12
0.40
Tript V2 discriminator timewalk
average discriminator firing time [nsec]
355
ch0
ch1
ch2
ch3
ch4
ch5
ch6
ch7
ch8
ch9
ch10
ch11
ch12
ch13
ch14
ch15
350
345
340
335
significant timewalk and chanto-chan spread for small signals
can set threshold at 1.5 p.e.
and discriminator will fire,
but timestamp for low amplitude
signals will not be reliable
OK for signals > ~ 3 p.e.
can correct for timewalk off-line
330
325
0.0
0.5
1.0
1.5
2.0
overall injected charge [pC]
1 2 3 p.e. (1 p.e. = 80 fC (5x105)
Mark Raymond - 27/6/06
13
TFB (Tript Front-end Board) prototype status
main functionality:
4 Tript’s/TFB => 64 SiPM channels (for ECAL)
individual programmable HVtrim (5 V range) for each SiPM channel
tript O/P signal digitisation
front end electronic calibration
FPGA to program tript, sequence operation, timestamp hits, control digitisation, format and transmit data, …
local LV power regulation
prefer to prototype designs for individual functions as much as poss. before committing to final TFB prototype
results here for on-board ADC, HVtrim and electronic calibration circuits
Mark Raymond - 27/6/06
14
prototyping elements of TFB
cal cct
SiPM
AD9201
Tript
HVtrimDAC
miniature coax and connectors
necessary to proove as much of TFB circuitry as possible before committing to layout
helps to identify where extra layout care is needed
improves chances of TFB prototype working successfully
Mark Raymond - 27/6/06
15
SiPM -> TFB connection - details
HVglobal
significant no. of
passives/channel –
need careful, high
density layout
47k
50V, 0402
220pF
50V
0402
47k
50V, 0402
cal test
pulse 10pF
100V, 0603
SiPM
coax sheath not DC
coupled to GND
1k
LV, 0402
trip-t
100pF
100V
0603
51R
LV
0603
100nF
LV
0402
330pF
100V
0603
10pF
100V
0603
HVtrim(0-5V)
HVglobal: common to all SiPM channels on TFB
HVtrim: individual for each SiPM channel, 5V adjustment range (choice of 8/10/12 bit DAC precision)
HVtrim applied to coax sheath – AC but not DC coupled to GND
Mark Raymond - 27/6/06
16
ADC for the TFB
AD9201 – used by D-zero
dual-channel => 2 tript’s/ADC
28 pin SSOP package
separate analog and digital supplies
5V analogue – needed to
accommodate tript O/P range
3.3 V digital
Mark Raymond - 27/6/06
17
tript linearity measured with AD9201
1000
analog supply and ADC reference
voltage configuration optimised so
that tript output signals well matched
to 10 bit ADC range
900
800
ADC units
700
600
500
400
300
high gain channel
low gain channel
200
100
0
0
5
10
15
20
25
30
35
40
Qin [pC]
Mark Raymond - 27/6/06
18
SiPM signals measured with tript/AD9201
Russian SiPM: gain 5.6x105
3
3000
14x10
with LED pulse
no LED pulse
2500
12
10
2000
8
1500
6
100,000 events in each spectrum
counts (no LED)
counts (with LED)
275 ns preamp integration period
~ 10 ADC units / p.e.
=> 0.1 p.e. ADC resolution
1000
4
500
2
0
0
140
160
180
200
220
240
260
ADC units
Mark Raymond - 27/6/06
19
HV trim circuit for TFB
HVglobal
51R
LV
0603
SiPM
coax sheath carries
HVtrim voltage
1k
LV, 0402
100nF
LV
0402
HVtrim(0-5V)
8 channel DAC chip => 2 / tript, 8 / TFB
8/10/12 bit versions available
identical chips, just different resolution
(price difference but negligible to us)
TSSOP 16 pin SM package
serial interface to program (from FE FPGA)
output voltage variable 0 -> 5 V (with 5 V power)
Mark Raymond - 27/6/06
20
TFB HV trim circuit linearity
5
8 bit DAC version used here
single DAC channel measurement
4
gives 20 mV precision for 5 V range
should be enough?
4
3
0
2
residuals [mV]
Output voltage
2
-2
1
-4
0
0
50
100
150
200
250
DAC value
Mark Raymond - 27/6/06
21
TFB HV trim circuit with SiPM
800
DAC value = 0
HVtrim = 0
HV eff. = 50 Volts
600
1600
1200
400
800
200
400
0
0
100
200
300
400
DAC value = 50
HVtrim = 1.0 Volts
HV eff. = 49 Volts
500
100
200
300
400
500
SiPM LED spectra for device with
nominal 47.5 V operating voltage
showing effect of HVtrim circuit
5 Volt range for HVtrim gives
overall range 45 – 50 Volts
(when combined with HVglobal)
8000
3000
DAC value = 100
HVtrim = 2.0 Volts
HV effective = 48 Volts
2000
4000
1000
2000
0
0
30
25
20
15
10
5
0
200
300
400
500
100
200
300
400
500
25
3
20
DAC value = 250
HVtrim = 4.9 Volts
HV effective = 45 Volts
15
x10
DAC value = 200
HVtrim = 3.9 Volts
HV effective = 46 Volts
x10
3
100
DAC value = 150
HVtrim = 2.9 Volts
HV effective = 47 Volts
6000
10
5
0
100
200
300
400
Mark Raymond - 27/6/06
500
100
200
300
400
500
22
CAL circuit for TFB
to 16 trip-t SiPM channels
before gain splitting capacitors
10 pF
4 CAL lines feeding
every 4th channel
Vcal (0 – 5 V)
(use another
AD5308 DAC here)
from FE-FPGA
discrete
MOSFETs
Mark Raymond - 27/6/06
23
CAL circuit test results with tript/AD9201
16 high gain chans
2.2
2.0
Volts
1.8
1.6
pedestal
DAC val. = 0
DAC val. = 5
DAC val. = 10
DAC val. = 15
DAC val. = 20
DAC val. = 25
tript multiplexed analog output stream
for different DAC values for one CAL
test input – sampled with scope
tript MUX (and ADC) running at 5 MHz
1.4
1.2
16 low gain chans
1.0
2.2
time [500 ns /division]
2.0
pedestal
DAC=50
DAC=100
DAC=150
DAC=200
DAC=250
Volts
1.8
1.6
1.4
substantial crosstalk – but only after
high gain channels beyond saturation
1.2
1.0
time [500 ns / division]
Mark Raymond - 27/6/06
24
CAL circuit test results with tript/AD9201
linearity measured for one SiPM channel
using external test pulse and CAL circuit
1000
-> close correspondence
800
ADC units
(also using AD9201)
600
400
high gain / external test pulse
low gain / external test pulse
high gain / CAL cct
low gain / CAL cct
200
0
0
10
20
30
40
Qin [pC]
Mark Raymond - 27/6/06
25
TFB elements prototyping summary
tript output ADC, SiPM HVtrim DAC circuit and electronic chain calibration circuit all prototyped
and tested
no major problems encountered
can now proceed to lay out the TFB prototype with confidence that at least these elements
should function OK.
Mark Raymond - 27/6/06
26
TFB layout status – 10 cm x 16 cm
HVtrim 16 SiPM I/Ps and passives HVtrim
CAL cct
Tript
AD9201
footprint
Mark Raymond - 27/6/06
FPGA footprint
high density SiPM I/P layout
complete – gives confidence that
size target ~ achievable
still much left to do (e.g. FPGA
dig. signals routing, power regs.,
connectors (power and control),
slow control interface, …..
27
TFB mounting ideas (ECAL)
TFB
cooled Al mounting plate
to SiPM
thermal gap filler
TFB mounted on cooled Al plate with cutouts through which SiPM cables are fed
min. coax connectors (and other connectors) on top surface
chips to be cooled on bottom surface, in thermal contact with plate
thermal gap filler allows for differences in chip thicknesses
power regs. on top side – dissipating heat to board – so will need to provide good thermal pathway
to mounting plate in this area of TFB
coax socket
~2 mm dia.
Mark Raymond - 27/6/06
terminated coax cable
(1.3 mm dia.)
28
TFB interfaces
4 LVDS pairs (RJ45 type connector and cable – should be screened)
Clocks input:
100 MHz, 1Hz, Spill/Cosmic trigger
Data input
Data output
RF clock ? (maybe not needed)
slow control
TBD (maybe just a connector to plug-on micro-controller based circuit?)
Power
< ~100V
+2.5
+5
+3.3
+1.2
Mark Raymond - 27/6/06
small
~ 0.5A
~ 0.2A
TBD
TBD
SiPM HV
tript and FPGA I/O
ADC analogue and HVtrim DAC
ADC digital and FPGA I/O
FPGA core
29
some data volume numbers
for programming
tript: ~ 900 kbits for 50k channels
HVtrim DACs: 8 bits res’n x 50k chans = 400 kbits
for raw spill data readout (data only)
assume
23 integration periods
4 tript’s / TFB
32 channels/tript (hi and logain)
10 bit ADC
=> ~30k bits /TFB /spill
+ hit timestamp data and associated hit channel addresses
Mark Raymond - 27/6/06
30
planning
Plans for this year (2006)
1st TFB prototype to be produced by end October
in parallel produce sufficient firmware for characterization
detailed electrical characterization by beginning 2007
Plans for next year (2007)
vertical slice test (1st quarter)
TFB prototype with photosensors, RMM and MCM prototypes
review requirements and design 2nd (final) TFB prototype for ECAL
produced and tested by end of year
Mark Raymond - 27/6/06
31