Silicon Strip Readout
and the
XYTER Electronics Development
Christian J. Schmidt et al.,GSI Darmstadt
10th CBM Collaboration Meeting, Dresden, Sept. 24. – 28., 2007
10th CBM Collaboration Meeting, Dresden Sept. 24 – 28, 2007
Challenging Experiments, Challenging Detector Specs
As a fixed target machine, very high event-rates (1000 tracks at 10MHz for CBM)
Up to 5% channel occupancy,
Very inhomogeneous track distribution (forward cone)
Signal latency will be too large for a timely, reasonable trigger decision,
Event overlap expected
Need front-end electronics:
MIPs at high input capacitance
highly integrated,
asynchronous, autonomous hit
detection,
time stamp labeling.
10th CBM Collaboration Meeting, Dresden Sept. 24 – 28, 2007
Additionally
Very harsh radiation environment (10 MRad)
Very high density channels with need for either
low radiation-length detector design (CBM-STS, PANDA TPC, PANDA forward GEM trackers, CBM TRD)
... or very compact detector design (e.g. CBM-Muon Chambers)
10th CBM Collaboration Meeting, Dresden Sept. 24 – 28, 2007
n-XYTER: Novel FE-Chip Architecture Cast in Silicon
Architectural Solution for FAIR CBM and PANDA.Starting point towards a FAIR dedicated XYTER front-end ASICOur work-horse readout ASIC for detector prototyping
detector readout ASIC for high-density and high statistical rate time and amplitude measurement
128 channels @ 50.7 µ pitch
freely running,
self triggered autonomous hit detection
850 (1000) ENC at 30 pF
dynamic range for 6 MIPs (300µ Si)
positive and negative signals
Per channel analogue energy and digital time stamp FIFO (1ns resolution)
De-randomizing, sparsifying Token Ring readout at 32 MHz
n-XYTER was
developed for neutron
applications within EU
FP-6 NMI3
10th CBM Collaboration Meeting, Dresden Sept. 24 – 28, 2007
chargepreamp
FASTshaper 18.5 ns
peaking
SLOW shaper(2 stages)
140 ns peaking time
Peakdetector &
hold,free running
comparatorTime Walk
Compensationcircuit
PDHreset
pulse height
output
triggertimestamp reg.
chargeinput
Data Driven Front-End: Asynchronous Channel Trigger
dig. FIFO
analogue FIFO
Asynchronous registry and storage in
4-level fifo guarantees data loss < 4 %
when read-out through token ring
The DETNI ASIC 1.0, a front-end evaluation chip in AMS 0.35µ
detection of statistical, poisson distributed signals
10th CBM Collaboration Meeting, Dresden Sept. 24 – 28, 2007
Analogue Signal Sequence (Test Channel)
Testpulse Release
Slow Shaper
Fast Shaper
Discriminator Output
10th CBM Collaboration Meeting, Dresden Sept. 24 – 28, 2007
Test Board for Tests on n-XYTER
64/128 chan. connected
I²C-Interface
Test points accessible
All functional tests possible
Digital output accessible
One additional analogue test channel is available for direct access of slow and fast shaper outputs... with output buffer would
have been even more useful
10th CBM Collaboration Meeting, Dresden Sept. 24 – 28, 2007
Analogue Pulses, Peaking Time, Front-End Noise
FAST channel SLOW channel
ENC26.9 e/pF + 200 e
12.7 e/pF + 233 epeaking
timea (1% to 99%)
18.5 ns 139 ns
Engineered for 30 pF, giving (850 ) 1000 e 600 e
pre-amp and shaper power consumption: 12.8 mW per channel; OK for neutrons!
10th CBM Collaboration Meeting, Dresden Sept. 24 – 28, 2007
Slow Shaper Output, the Energy Channel
-250 0 250 500 750 1000 1250 15000,5
0,6
0,7
0,8
0,9
1,0
1,1
1,2
1,3
slow
sha
per
outp
ut (
V)
t in ns
20 fC 6,6 fC 3,1 fC
detector cap 10,2 pF
-1000 -500 0 500 1000 1500
-0,40
-0,35
-0,30
-0,25
Slo
w S
hape
r O
utpu
t (V
)
t in ns
4,9 pF 10,2 pF 21,2 pF 99,2 pF
Measurements on the test channel #129
varying input charge
varying input capacitance
10th CBM Collaboration Meeting, Dresden Sept. 24 – 28, 2007
Some Inter-Channel Pick-Up, Ongoing Detective Work
Cin = 0 pF, bond wire removed
System Effect
No dependence upon no. of bond wires, power or gnd
May be worsened with discriminator settings (TWC)
Cin = 22 pF
Feedback via spurious coupling through epitaxial, optical layer,
substrate to the input
10th CBM Collaboration Meeting, Dresden Sept. 24 – 28, 2007
Some In-Channel Discriminator Feedback Detected
...upon removal of discriminator-power decoupling
These issues are particularly important with the self triggered architecture!
They will be addressed even more in the next engineering run.
correlates with internal discriminator trigger
correlates with external test-pulse release signal (blue)
10th CBM Collaboration Meeting, Dresden Sept. 24 – 28, 2007
Channel layout overview, Clock Domains and Power
analogue front end , PDHcomp – TWC
trim regmaskreg.
analoguemem.
ch.ID
tokencell
TSlatch
digitalmem.
monosynchcontrol
input MOSGND
analogueGND
& BULK
analogueVDD
A/Dguard ring
digitalBULK
digitalGND
digitalVDD
clocktree
comparatorVDD
memory control ( 9 bit )PDH reset
comp
PAD
analogue domainno clock
digital domainsystem clock
8 mm
time stampfast clock
5 mm
total of 4 nF on chip MIM caps
10th CBM Collaboration Meeting, Dresden Sept. 24 – 28, 2007
Token Ring Readout, Data Transmission
-100n 0 100n 200n 300n 400n 500n 600n
Time (s)
Input Trg
CLK (128) CLK (32)
D7 D6 D5 D4 D3 D2 D1 D0
Data Valid
TS grey codedCh# grey coded
data transfer tested at 35 MHz, will also work at 128 MHz
Measurement: Transmission of four data elements on channels 1, 8, 30, 82
10th CBM Collaboration Meeting, Dresden Sept. 24 – 28, 2007
Analogue Differential Output, the Energy Channel
-1000 0 1000 2000 3000 4000 5000-0,30
-0,25
-0,20
-0,15
-0,10
-0,05
0,00
0,05
0,10
0,15
Diff
eren
tial O
utpu
t B
uffe
r S
igna
l (V
)
Time in ns
-10 -5 0 5 10 15 20
-0,16
-0,14
-0,12
-0,10
-0,08
-0,06
-0,04
-0,02
An
alo
g D
iffe
ren
tial O
utp
ut (
V)
t in ns
Signal settling upon successive data
Three signals, one signal altered
10th CBM Collaboration Meeting, Dresden Sept. 24 – 28, 2007
Power Consumption
preamplifier 7.4mW
fast shaper 2.5mW
slow shaper stage 1 1.7 mW
slow shaper stage 2 2.5 mW
discriminator 2.1 mW
peak detector and hold 2.7 mW
analogue FIFO 2.3 mW
overall we find 21 mW/channel operating power
10th CBM Collaboration Meeting, Dresden Sept. 24 – 28, 2007
Testing Summary
Slow Control Operative
Clocking at 256 MHz possible
Grey coded time stamp generated
Analogue readout operative
Features operative: Positive and negative signal processing
Global threshold, local threshold fine tune
On chip test pulse generation
Channel masking
Channel forced trigger (baseline determination)
Individual channel shut down
Pile-up lableing, fifo overflow identification
Testers involved:
Gerd Modzel (PI HD)
Markus Höhl (GSI)
Knut Solvag (GSI)
Sharma Anurag (Dehli)
Rafal Lalik (AGH, GSI)
Adam Czermak (AGH)
thanks for his support to
Sven Loechner (GSI) and others
10th CBM Collaboration Meeting, Dresden Sept. 24 – 28, 2007
n-XYTER Engineering Run in Preparation
Engineering Run Targeted for Jan. 2008
H. K. Soltveit (PI Heidelberg)
Addresses active feedback baseline adjustment
full dynamic range, no temp. co. on base-line
Addresses spurious feedback coupling through substrate
Reduction of power consumption where easily possible
From 250 Dies to Work Horse Electronics for Detector Prototyping
Will yield several thousand chips for prototyping of
CBM STS, PANDA TPC and other detectors
as well as the DAQ chain and beyond.
Go
Detector
Prototyping
10th CBM Collaboration Meeting, Dresden Sept. 24 – 28, 2007
Dedicated XYTER Development for FAIR
Generic n-XYTER architecture finds broad applications within FAIR:
CBM Silicon Strips STS, High Rate GEM TPC as well as large area
gas detectors (micro structures or wire chambers)
Twin chip development with XYTER architecture and diversities for:
• Radiation hard design in UMC 0,180 µm (better than 10 MRad)• Minimized power consumption • Integration of modern, low power ADC on chip purely digital interface• Highly multiplexed data interface (minimize cableing)• Optimized system synchronization capabilities• SEU tolerance• Detector DC coupling capability• Dense mounting capability
Silicon Strip MIP detection,
micro-structured gas detectors
Larger dynamic range,
ion tail cancellation specialties
10th CBM Collaboration Meeting, Dresden Sept. 24 – 28, 2007
Groups Involved In XYTER Development
AGH Krakow, (Robert Szczygiel, Pawel Grybos et al.)
TI Heidelberg (Peter Fischer and Tim Armbruster)
MEPHI, Moscow (E. Atkin et al)
PI Heidelberg and HD ASIC-Lab (H. K. Soltveit)
GSI Darmstadt, C.J. Schmidt (coordination) and Sven Loechner
Further engaged:
S. Chattopadhyay et al, Kolkatta on individual design items
University Bergen, K. Ulaland and D. Röhrig et al
We will structure the XYTER development collaboration Friday morning 9:00 to 11:00
CBM
10th CBM Collaboration Meeting, Dresden Sept. 24 – 28, 2007
Thank you for your attention
CBM
10th CBM Collaboration Meeting, Dresden Sept. 24 – 28, 2007
Token Ring Architectural Pros/Cons
High Efficiency Empty channels automatically skipped in readout process Built-in fair distribution of readout bandwidth, automatic bandwidth focussing
Built-in De-Randomization: 100% bandwidth used on data
Error Robustness
Any problematic channel (e.g. continuously firering) will divert and occupy a
maximum of 1/nth of the bandwidth.
Built-in, non-perfect readout probability avoids unrecoverable logic deadlock:
Problematic situations like any kind of pile-up, logic hang-ups or glitch
cause mere deadtime but the “show will go on”.
But: Data needs to be tagged with a time-stampData needs to be resorted and re-bunched after readout
10th CBM Collaboration Meeting, Dresden Sept. 24 – 28, 2007
Investigating Individual Channels, Triggerefficiency
Trigger efficiency in Treshold Scan: The S-Curves
- Input of test pulses at fixed rate,- scan threshold while measuring detection rate
Derivative gives image of noise!
10th CBM Collaboration Meeting, Dresden Sept. 24 – 28, 2007
Trigger efficiency tested for all channels
origin of 2 ch periodicity attributed to four fold pulser circuitry
Channel number
note bonded and non-bonded channels
10th CBM Collaboration Meeting, Dresden Sept. 24 – 28, 2007
Token Ring Readout Process
Focus bandwidth where there is data
32 MHz data readout Automatic zero
suppression (sparsification)
Analog FIFO
Timestamp FIFO
datareadout
bus
token cycle
Disc.
token cell control logic for
data readout or token passskip channels without data, asynchronously rush through empty channels until data found
10th CBM Collaboration Meeting, Dresden Sept. 24 – 28, 2007
Modelling FIFO Occupancy
FIFO can virtually be filled with up to 2*n events with no data losssince n elements are read while data comes in.
1 2 3 4 5 6 7 8 9 10 11 1200
0.05
0.1
0.15
0.21 2 3 4 5 6 7 8 9 10 11 12 13
We loose from here on
Pro
bab i
li ty
Poisson Distribution(): e.g.: fifo depth n = 4, so expect 4 events during readout if incoming rate equals maximum readout rate. = 4.
10th CBM Collaboration Meeting, Dresden Sept. 24 – 28, 2007
n-XYTER Front-End Topology