twepp - 11 – 26-30 september 2011 – vienna, austria
DESCRIPTION
Low Noise Front End ASIC with Current Mode Active Cooled Termination for the Upgrade of the LHCb Calorimeter. E. Picatoste, C . Abellan, O. Duarte, L. Garrido, D. Gascon, J. Lefrançois, E. Grauges, F. Machefert, X. Vilasis Universitat de Barcelona Institut de Ciències del Cosmos ICC-UB. - PowerPoint PPT PresentationTRANSCRIPT
TWEPP-11 – 26-30 September 2011 – Vienna, Austria
Low Noise Front End ASIC with Current Mode Active
Cooled Termination for the Upgrade of the LHCb
CalorimeterE. Picatoste, C. Abellan, O. Duarte, L. Garrido, D. Gascon, J.
Lefrançois, E. Grauges, F. Machefert, X. Vilasis
Universitat de Barcelona
Institut de Ciències del Cosmos ICC-UB
2
Outline
1. Introduction
2. Channel Architecture
3. First prototype
4. Second prototype
5. Summary
27th September 2011 TWEPP-11
TWEPP-11 3
LHCb Upgrade
• In ~5 years of data taking, most of the proposed physics program will be covered.
• In order to distinguish between models of new physics, a large increase in data rate is required.
• LHCb Collaboration intends to upgrade the detector during the long LHC shutdown in 2016.
OT
VELO
RICH-1
TT
ITMagnet
Muon StationsM2-M5
-5m
5m
y
10m 0m
Calorimeters
RICH-2
M1
27th September 2011
4
Current ECAL/HCAL FE
AD
C
L0
Pipeline
ECAL
HCAL sameelectronics
PM
PM
Analogue 10 mclip
clip
Readout
Trigger
•Pulse shaping in 25 ns•Residue < 1% after 25 ns •Integrator plateau : 4 ns•Linearity < 0.5%•Rise time ~ 5 ns
LAL-OrsayLAL-Orsay
TWEPP-1127th September 2011
5
Motivation
• PM current has to be reduced:– Otherwise PMT would die rapidly (require a factor ~ 5)– FE electronics gain has to be increased correspondingly– FE noise should not be increased in the operation!
• New front end board is required:– Low noise analog electronics– GBT for data transmission @ 40MHz
• For 12 bit DR, input referred noise:– Voltage amplifier: < 1 nV/sqrt(Hz)– Current amplifier: < 10 pA/sqrt(Hz)– Active cooled termination required
• But 2/3 of the signal are lost by clipping:– Alternative solution: remove clipping at the PM base (detector)– Perform clipping after amplification in FE– Alternative discrete elements + delay line solution
TWEPP-1127th September 2011
6
Channel Architecture
• Current mode amplifier• Switched integrator
– Fully differential Op Amp
• Track and Hold– 12 bit: flip-around architecture
• Analog multiplexer• ADC driver
– To match ADC input impedanceMP2MP1
Q1
Re
Ib1 MN2
MP4MP3
MN1
Rf
Vee
Vcc
1 : K : mK : mK
Ib2
Ii
Io ∫
∫
I I
Current amplifier (mirrors)
TH
TH
AnalogueMultiplexer
Drv
ADC driver
Very difficult to integrate HQ analog
delay lines
2 switched alternated paths as in PS/SPD (LHCb
CALO)
Switching noise!
Fully diferential ASAP
TWEPP-1127th September 2011
7
Channel Architecture: Current Amplifier Input Stage
• Current mode feedback: – Inner loop: lower input impedance
• Current feedback (gain): mirror: K
– Outer loop: control input impedance
• Current feedback: mirror: m
• Current gain: m• Electronically cooled input
impedance
• Current mode feedback used – Optical comunications– SiPMreadout
• Low voltage– Only 1 Vbe for the super common
base input stage
• Better in terms of ESD:– No input pad connected to any
transistor gate or base
fem
i mRK
K
K
RgZ
11
1 1
MP2MP1
Q1
Re
Ib1 MN1
MP4MP3
MN2
Rf
Vcc
1 : K : K : nK : nK
Ib2
Ii+
MP5
MN3 MN4
m : 1 : n : n
Vref
Io+_Sucbh1 Io+_Sucbh2
Io-_Sucbh1 Io-_Sucbh2
Inner loopOuter loop
2 differential outputs
TWEPP-1127th September 2011
8
Channel Architecture: Switched Integrator
• Switched integrator– Integrator plateau : 4 ns
– Linearity < 1%
– Residue < 1% after 25 ns
– Reset time < 5 ns
+
-
-
+
In+
In-Out+
Out-
FDOA
CMOS switches
• Rf to improve plateau- 100 ns time constant
- Cancel slow component of the clipped signal (remanent)
-1,2
-1,0
-0,8
-0,6
-0,4
-0,2
0,0
0,2
0 25 50 75 100 125 150 175 200
t (ns)
V n
orm
aliz
ed
Clipped Pulse (Anatoli)
Test Pulse
TWEPP-1127th September 2011
9
Channel Architecture: Track-And-Hold
• Track-And-Hold:- 12 bit: flip-around architecture
Vin+
Vin-
Vrefth
clkd
clkd
clk
clk_n
clk_n
clkFDOA
-
+
+
-
clk
Vo-
Vo+
TWEPP-1127th September 2011
10
Channel Architecture: FDOA
• Fully differential Op Amp:- Folded cascode + Miller stage with CMFB
M
M
Vin+
M
Vcas
VrefCE
Vout+
M
M
M
M
Vin-
M
Vcas
VrefCE
Vout-
M
MM
M
Vcmref
M
M
Vb1
Vb2 Vb2Vb2
Vb1
Vb2
Folded cascode
NPN CE amp
Pole compensation
RDegenertion
Common Mode Feedback
FDOA specifications
Parameter Value
Gain bandwidth 500 MHz
Phase margin > 65º
Slew rate > 0.4 V/ns
VCM 1.65 V
TWEPP-1127th September 2011
11
Choice of Technology
TWEPP-11
SiGe BiCMOS is preferred: SiGe HBTs have higher gm/Ibias than MOS: less noise, less Zi variation SiGe HBTs have higher ft (>50 GHz): easier to design high GBW amplifiers
Several technologies available: IBM IHP AMS BiCMOS 0.35 um
AMS is preferred Factor 2 or 3 cheaper Too deep submicron CMOS not required / not wanted:
Few channels per chip (4 ?) Smaller supply voltage Worst matching
Radiation hardness seems to be high enough 30 Krad seems to be OK (checked by other Upgrade projects)
IBM IHP AMS
HBT ft > 100 GHz 190 GHz 60 GHz
CMOS 0.13 um 0.13 um 0.35 um
Proto Cost [€ /mm2]
> 3 K > 3 K 1 K
27th September 2011
12
First prototype
Switched integrator
Track and Hold
MP2MP1
Q1
Re
Ib1 MN2
MP4MP3
MN1
Rf
Vee
Vcc
1 : K : mK : mK
Ib2
Ii
Io ∫
∫
I I
Current amplifier (mirrors)
TH
TH
AnalogueMultiplexer
Drv
ADC driver
ICECAL chip:SiGe BiCMOS 0.35umAMS 2 mm2
Received: October 201012 chips
• Purpose: test key points of a novel circuit idea– Input impedance control by current feedback– Low noise performance– Dynamic range: Linearity
• Also, to test critical aspects of a switched solution:– Offset between subchannels– Noise– Plateau of the integrator output
• Effect of the clock jitter versus signal
TWEPP-1127th September 2011
13
SCOPE
AWG2021
50 Ohms50
cm
10m
IN OUT
ICECALDiff
eren
tial
Pro
be
Differential Probe
50 Ohms
First Prototype: Input Impedance
IC in
put s
igna
l
(PC
B)Integrator
clock
Waveform Generator
Source signal
sbch
1sb
ch2
Reflection Coefficients for source, IC input, and output signals:
Refl. Coeff. = 1st pulse integral / 2nd pulse integral
TWEPP-1127th September 2011
14
SCOPE
AWG2021
50 Ohms50
cm
10m
IN OUT
ICECALDiff
eren
tial
Pro
be
Differential Probe
50 Ohms
Input impedance
IC in
put s
igna
l
(PC
B)
Waveform Generator
Source signal
sbch
1sb
ch2
ICECAL OFF
Large reflections
Integrator clock
TWEPP-1127th September 2011
15
Input impedance
AWG signal
50 470
1uF
1uF
ICECALIN
sbch1
sbch2
INT
INT_BAR
CLOCKPCB
Rp
• PCB input circuit with AC coupling with Rp:– Zin a little higher than 50Ω – Rp used to fine tune Zin– Optimal Rp is between 360 and 390 Ω
• Dynamic variation of input impedance is << 1 % for full dynamic (50 pC)
• Measurement error for second reflection is quite high for low amplitudes (diff probe noise)
• Low Refl. Coef. dispersion < 0.5%
Impedance adaptation
-5
-4
-3
-2
-1
0
1
2
3
4
5
0,0E+00 1,0E-11 2,0E-11 3,0E-11 4,0E-11 5,0E-11 6,0E-11 7,0E-11
PM delivered charge [C]
Area
Coe
ffici
et [%
]
First Reflection
Second reflection
Integrator output
Rp = 390 Ohm
Histogram of input impedance matching for 12 chips
0
2
4
6
8
10
12
14
16
18
-2,5 -2 -1,5 -1 -0,5 0 0,5 1 1,5 2 2,5
Area coefficient [%]
Entr
ies
First
Second
Int Output
-1.75±0.43%
0.82±0.11%
1.99±0.21%
TWEPP-1127th September 2011
16
First Prototype: Noise
• The noise:– Generate a histogram of the voltage value by the end of the
cycle (from 100 waveforms).
– Gaussian fit => standard deviation
• Correlated Double Sampling (CDS) ↔ dynamic pedestal subtraction:
– LF pick-up noise at measurements (chip connection/socket)
– without CDS: 1.8 ADC counts (LF noise!)
– with CDS: 1.2 ADC counts (≈1 ADC count from simulations)
Integrator output
-6,00E-02
-4,00E-02
-2,00E-02
0,00E+00
2,00E-02
4,00E-02
6,00E-02
8,00E-02
1,00E-01
0,00E+00 2,00E-08 4,00E-08 6,00E-08 8,00E-08 1,00E-07 1,20E-07
Time [s]
[V]
Output fluctuation at different sampling times
0
10
20
30
40
50
60
-10 -8 -7 -6 -5 -4 -3 -2 -1 0 1 2 3 4 5 6 7 8 9 10
ADC counts
Entr
ies
Value @ T0
Value @ T0-50 ns
Value @ T0 - 100ns
Output fluctuation sampled with dynamic pedestal subtraction
05
101520253035404550
-10 -8 -7 -6 -5 -4 -3 -2 -1 0 1 2 3 4 5 6 7 8 9 10
ADC counts
Entr
ies
CDS 50n
CDS 100n
TWEPP-1127th September 2011
17
First Prototype: Offset
Offset measurement:V at cycle end
-2ns
V offset measurement
Preamp current offset is integrated
=> Offset at output
Distribution of offset difference between subchannels
0
0,5
1
1,5
2
2,5
3
3,5
-100 -80 -60 -40 -20 0 20 40 60 80 100
Offset difference [mV]
Entr
ies
Diff Subch1-Subch2
ABS (Diff Subch1-Subch2)
5 % of dynamics
• With AC coupling between ICECAL and ADC what really matters is the difference between the offset of the 2 subchannels
• Well below the 5 % of the full scale range (2 V)
TWEPP-1127th September 2011
18
First Prototype: Linearity
• Relative linearity error < 1% for the full range (2V)
• Error is large for very low amplitude values
• Attenuator + WaveCatcher offer better linearity than the oscilloscope but the differential probe may need calibration
-5
-4
-3
-2
-1
0
1
2
3
4
5
0 0,5 1 1,5 2 2,5Vout (V)
Lini
eari
ty E
rror
(%)
TWEPP-1127th September 2011
19
Plateau at the integrator output
• Due to clock jitter, the signal at the output must be stable (<1%) for 4 ns.
• Input signal AWG generated similar to clipped.
• Method:– Delay clock signal in 1ns increments– Use LEMO cable
Effect of time alingment (signal versus clock)
-5
-4
-3
-2
-1
0
1
2
3
4
5
-10 -5 0 5 10
Delay variation [ns]
Re
lati
ve
va
ria
tio
n [
%]
150 mV output600 mV output1.2 V output
• The output variation is smaller than 1% for about 3 ns delay variation
– Consistent for different signal amplitudes
• Can be improved using a resistor in parallel to the integrator capacitor (ICECAL2)
+
-
-
+
CF
CF
In+
In-Out+
Out-
RF
RF
TWEPP-1127th September 2011
20
Second prototype
Switched integrator
Track and Hold
MP2MP1
Q1
Re
Ib1 MN2
MP4MP3
MN1
Rf
Vee
Vcc
1 : K : mK : mK
Ib2
Ii
Io ∫
∫
I I
Current amplifier (mirrors)
TH
TH
AnalogueMultiplexer
Drv
ADC driver
ICECAL2 chip:SiGe BiCMOS 0.35umAMS 2 mm2
Submitted: June 6th 2011To be received in September
• Purpose: test key points of the circuit– Input impedance control by current feedback– Low noise performance– Dynamic range: Linearity
• Also, to test critical aspects of a switched solution:– Offset between subchannels– Noise– Plateau of the integrator output
• Effect of the clock jitter versus signal• Integrator Rf effect
TWEPP-1127th September 2011
21
Second Prototype
Clock input
Input signal
Integrator output
TH output
TWEPP-11
Main signal Tail
Main signal
Tail
27th September 2011
22
Second Prototype Simulations
TWEPP-11
• Input impedance is the same as the 1st prototype
• Linearity:– After TH linearity < 1% for output
amplitudes < 1.8V– The linearity is limited by the
output stage NMOS of the FDOA• Apply offset to the
differential signal so the output voltage is reduced
• The noise:– Generate a histogram of the
voltage value by the end of the cycle (from 100 waveforms).
– Zero input signal– Gaussian fit => standard deviation– Noise 0.8 ADC counts≲
27th September 2011
23
• An integrated circuit for the Upgrade of the LHCb Calorimeter Front End electronics has been presented. Architecture based on:
Current mode preamplifier with cooled termination for reduced input noise 2 fully differential interleaved channels Switched integrator Track-and-Hold
• Measurements of the first prototype of the input stage (preamp+integrator) are about finished Principle is ok Good results with 12 chips statistics Need to study the effect of bias (op. paint) variation
• A second prototype was submitted in June It includes: preamp + integrator + track-and-hold Added a feedback resistor to improve the integrator output stability
Summary
TWEPP-1127th September 2011
24
Back-up:
TWEPP-1127th September 2011
25
Test set-up
Waveform Generator signal vs. Measured clipped signal (Anatoli)
-1,2
-1,0
-0,8
-0,6
-0,4
-0,2
0,0
0,2
0 25 50 75 100 125 150 175 200
t (ns)
V n
orm
aliz
ed
Clipped Pulse (Anatoli)
Test Pulse -1,2
-1,0
-0,8
-0,6
-0,4
-0,2
0,0
0,2
0 25 50 75 100 125 150 175 200
t (ns)
V n
orm
aliz
ed
Clipped Pulse (Anatoli)
Test Pulse
Fast pulse (no tail) to measure reflections
TWEPP-1127th September 2011
26
SCOPE
AWG2021 50 Ohms
50
cm
10m
IN OUT
ICECALDiff
ere
ntia
l P
robe
Differential Probe
50 Ohms
Test set-up
Tektronix AWG2021 Waveform Generator:
•250 MS/s•12 bit
50Ω || 50Ω = 25Ω emulate clipping impedance
Tektronix TDS7154B Oscilloscope:
•BW = 1.7 GHz•20 GS/s•8 bit
Differential probe:•BW = 400 MHz•Cin = 1 pF
TWEPP-1127th September 2011
27
Test set-up: linearity
AWG2021
50cm
10m
IN OUT
ICECALProbe Probe
50 Ohms
ATTENUATOR
Signal
CLK
Trigger
WaveCatcherCH0
CH1
PC
Attenuator: Better linearity
than AWG (<1%) No need for
timing corrections
WaveCatcher: Better linearity
than oscilloscope
PC: Automatized
systemTWEPP-1127th September 2011
28
Test set-up
TWEPP-1127th September 2011