Transcript vherrero_RT2010_v97

AMIC: AN EXPANDABLE FRONT-END FOR
GAMMA-RAY DETECTORS WITH LIGHT
DISTRIBUTION ANALYSIS CAPABILITIES
Vicente Herrero* , Christoph W. Lerche
Michelle Spaggiari, Ramón Aliaga,
Néstor Ferrando and Ricardo Colom.
17th REAL TIME CONFERENCE / IEEE - NPSS
Continuous Scintillator  Ray Detector
Designed for PET applications
LSO :
42x42x10 mm
q
PMT :
Hamamatsu
H8500

Lower cost.

Higher detector sensibility. [1]



Better energy resolution.
[2] P. Bruyndonckx, et al. “Initial Characterization of a
Nonpixelated Scintillator Detector in a PET Prototype
Demonstrator”, IEEE Trans. Nucl. Sci., (53), 2543, 2006.
[2]
Light Distribution Analysis
[1] P. Bruyndonckx, et al. “Performance Study of a PET
Detector Module Based on a Continuous Scintillator”,
IEEE Trans. Nucl. Sci., (53), 2536, 2006.
[3][4]
Geometry and Coating can be
changed for optimization. [5]
[3] C. W. Lerche, et al. “Depth of gamma-ray interaction
within continuous crystals from the width of its scintillation
light-distribution” IEEE Trans. Nucl. Sci., (52), 560, 2005.
[4] C. W. Lerche, et al. “Fast circuit topology for spatial
signal distribution analysis and its application to nuclear
medicine imaging” IEEE NPSS RT 2010,
[5] C. W. Lerche, et al. “Dependency of Energy, Position
and Depth of Interaction on Scintillation Crystal Coating and
Geometry", IEEE Trans. Nucl. Sci., 55, (2008) 1344.
01/04/2016
2
17th REAL TIME CONFERENCE / IEEE - NPSS
PESIC
Previous Work
◦ Individual Anode Gain Adjustment
for Detector Equalization [7]
◦ Reduces error due to signal delay
in resistor network [6]
[6] V. Herrero, et al. “PESIC: an integrated front-end for PET
applications”, IEEE TNS., (55), 27, 2008.
◦ Not an expandable architecture
◦ Low resolution in Depth of
Interaction (DOI) measurements [7]
[7] V. Herrero, et al. “Position sensitive scintillator based
detector improvements by means of an integrated frontend”,
NIMA., (604), 77, 2009.
01/04/2016
3
17th REAL TIME CONFERENCE / IEEE - NPSS
Mathematical Foundations of AMIC
Moment n of distribution f(x)
mx0   (1) f x ( x) x
ENERGY
m1x   x f x ( x) x
CENTROID (x AXIS)
mx2   x 2 f x ( x) x
VARIANCE (width of
distribution)
mx3   x 3 f x ( x) x
Kn(x)
…
SKEWNESS (assymetry
of distribution)
DISCRETIZED
BASIC
OPERATION
fx(0)
Kn(0)
fx(1)
Kn(1)
fx(2)
Kn(2)
fx(3)
Kn(3)
…
+
m
n
x
[4] C. W. Lerche, et al. “Fast
circuit topology for spatial signal
distribution analysis and its
application to nuclear medicine
imaging” IEEE NPSS RT 2010,
01/04/2016
4
17th REAL TIME CONFERENCE / IEEE - NPSS
AMIC Architecture
64 light
distribution
samples
I2C interface
(coef.
program.)
64 input
buffers
8 Computational
Blocks
8 Output Amplifiers
(current & voltage)
01/04/2016
5
17th REAL TIME CONFERENCE / IEEE - NPSS
Preamplifiers and Output Stage
Both SiPM and PMT
capable


Optimized PMT


Bandwidth > 34 MHz

Noise  0.1 uArms

40ux100u
Solved by
calibration of
coeff. values

Mismatch in current
mirror between
different preamps
Current Output
Voltage Output using
an R amplifier

Bandwidth > 20 MHz

SlewRate = 350 V/us
(CL= 50pF)

THD < 0.35 %
(0.5mApp)
175u x135u
01/04/2016
6
17th REAL TIME CONFERENCE / IEEE - NPSS
Computational Block
Analog Current Mode Filter
1540u x120u
Coefficient
Unit

Coefficient values < 1 with 8 bits precision

offers 256 different values for coefficients.



Vout=Vdump
40u x50u
[5] K. Bult and G. Geelen “An inherently linear and
compact most-only current division technique", 39th IEEEISSCC 189 (1992).
A linear distribution (M1) of 28 values on
one axis means  256x256 inputs
64*8=512 Coefficient units !!  Area needs
to be restricted
Extremely sensitive to Vout = Vdump
◦


Small differences introduce big linearity errors
Area ratio optimized for a maximum input
current of 6mA (BW > 60 MHz)
Needs biasing to get a better linearity
01/04/2016
7
17th REAL TIME CONFERENCE / IEEE - NPSS
Computational Block
200u x50u
Fully Differential
Current Collector

Necessary to stabilize DUMP & OUT
voltages to the same value

High bandwidth (>200 MHz), PM (>80º) and
Low THD (<0.05 %)

Matched layout needed for DUMP & OUT
current paths
◦
Differences in parasitic resistance introduce voltage
variations in DUMP & OUT voltages close to coeff.
◦
1 collector handles 16 coeff.
01/04/2016
8
17th REAL TIME CONFERENCE / IEEE - NPSS
Architecture Expansion using AMIC
8x
AMIC 1

AMIC II
8x

8x

AMIC is the basic building
block
Each AMIC generates up to 8
«partial moments»
1 AMIC is enough to
implement the final addition
of all «partial moments»
(up to 64 «partial moments» of 64
inputs which means 4096 inputs !!)
8x
AMIC III
AMIC IV
AMIC V

Anyway if you need more…
just add more AMICs at the
output
◦
Only noise generated in the front-end
limits the number of inputs
8x
01/04/2016
9
17th REAL TIME CONFERENCE / IEEE - NPSS
Performance Test Measurements
Linearity of
Coefficient Values
1
y(x) = a (x - b)
0.9
a = 0.0039294
b = 3.0374
0.8
R = 0.99991 (lin)
Normalized Value
0.7
0.6
0.5
0.4
0.3
0.2
0.1
0
0
50
100
150
200
250
Coefficient Code
Bandwidth
Switch matrix made with 8
ADG2128 (Analog Devices)
16 MHz
Linearity vs.
Input Current
(Coeff. Value
=128)
± 0.65 %
Input Dynamic
Range
(within Linearity
limits)
1.65 mA
THD
0.5 mA / 100kHz
0.5%
Noise
(current output)
1.8 µArms
Most codes lie
inside the limits
but…
7 effective bits
Area constrains in
Coeff. Unit introduce
mismatch
01/04/2016
10
17th REAL TIME CONFERENCE / IEEE - NPSS
Experimental Setup Test Measurements
5x5 Sweep on
detector surface
(6 mm steps)
01/04/2016
11
17th REAL TIME CONFERENCE / IEEE - NPSS
Experimental Setup Test Measurements

AMIC calibrated for preamp
mismatch compensation

Collimation Spot = 1 mm

M1x & M1y OUTPUTs (same



results)
No improvements made by
adding information of other
moments
As usual border effect
increases resolution far from
center
Just testing everything works
M1x calibrated coefficients
01/04/2016
12
17th REAL TIME CONFERENCE / IEEE - NPSS
Experimental Setup Test Measurements




AMIC calibrated for preamp
mismatch compensation
«Non invasive» method
DOI can improve 2D position
resolution and also reduce
parallax error after
reconstruction stage
Improvement of 300% over
previous results with PESIC
[7] V. Herrero, et al. “Position sensitive scintillator based
detector improvements by means of an integrated frontend”,
NIMA., (604), 77, 2009.
 of light distribution
+++
[3]
M2 calibrated coefficients
Depth of Interaction
[3] C. W. Lerche, et al. “Depth of gamma-ray interaction within continuous crystals from the width of its scintillation
light-distribution” IEEE Trans. Nucl. Sci., (52), 560, 2005.
01/04/2016
13
17th REAL TIME CONFERENCE / IEEE - NPSS
This is just the beginning…
The higher the number of inputs the better the
spatial resolution… but also:
 2D position calculation can be improved by
using other moments information
 DOI calculation can also be improved the
same way
 NOW WORKING WITH NEURAL NETWORKS
to optimize full reconstruction of detected events

01/04/2016
14
17th REAL TIME CONFERENCE / IEEE - NPSS
A lot of information waiting there
to be used …
Check out Christoph W. Lerche’s
work at the poster session !!!!
[4] C. W. Lerche, et al. “Fast
circuit topology for spatial signal
distribution analysis and its
application to nuclear medicine
imaging” IEEE NPSS RT 2010,
01/04/2016
15
17th REAL TIME CONFERENCE / IEEE - NPSS
Other work being carried out…

We also work on:
◦ Increasing Timing resolution in Coincidence Detection
 J. M. Monzó
◦ High Performance Data Acquisition Systems
 R. Aliaga
01/04/2016
16
17th REAL TIME CONFERENCE / IEEE - NPSS
THANK YOU FOR YOUR
ATTENTION
01/04/2016
17