Construction and Performance of a Double-Sided Silicon Detector Module using the Origami Concept C. Irmler, M. Friedl, M. Pernicka HEPHY Vienna

Construction and Performance of a Double-Sided Silicon Detector Module

using the Origami Concept

C. Irmler, M. Friedl, M. Pernicka

HEPHY Vienna

2TWEPP 2009, Christian Irmler (HEPHY Vienna)23 September 2009

Contents

Motivation

“Origami” Chip-on-Sensor Concept

Prototype Assembly

Beam Test Performance

Summary & Outlook


Motivation


Prototype Assembly


Summary & Outlook


~1 km in diameter

Mt. Tsukuba

KEKB Belle

KEKB and Belle @ KEK (Tsukuba, Japan)Asymmetric machine:8 GeV e- on 3.5 GeV e+

• e+/e- linear accelerator + storage ring

• Center of mass energy: Y(4S) (10.58 GeV)

• High intensity beams (1.6 A & 1.3 A)

Linac

Belle

KEKB

http://www.dex.ne.jp/mantan/search/std2_search_preview.jhtml?start=39&lastServiceTime=1091629566859&number=37


Belle-II

• KEKB factory upgrade until 2013• Target luminosity of 8 x 1035 cm-2s-1

• 40 times the present value• Accordingly increase of background

• Limitations of current silicon vertex detector (SVD2):– Occupancy need faster shaping– Trigger rate (dead time) need faster readout and pipeline

• Replacement of the SVD and its readout is necessary• APV25 readout chip would fit the needs


Comparison VA1TA – APV25

VA1TA (SVD2)• Commercial product (IDEAS)

• Tp = 800 ns (300 ns – 1000 ns)• no pipeline• 5 MHz readout• 20 Mrad radiation tolerance• noise: ENC = 180 e + 7.5 e/pF• time over threshold: ~2000 ns• single sample per trigger

APV25 (Belle-II SVD)• Developed for CMS by IC

London and RAL• Tp = 50 ns (30 ns – 200 ns)• 192 cells analog pipeline• 40 MHz readout• >100 Mrad radiation tolerance• noise: ENC = 250 e + 36 e/pF• time over threshold: ~160 ns• multiple samples per trigger

possible (Multi-Peak-Mode)

Must minimizecapacitive load !!!


Shaping Time and Occupancy

Threshold

Threshold

Tim e over threshold ~ 2000ns (m easured)

Tim e over threshold ~ 160ns (measured)

Sensitive tim e window ~ 20ns

VA1TATp~800ns

APV25Tp~50ns

Pulse shapeprocessingRM S(tm ax)~3ns

Gain ~12.5

Gain ~8

Total gain ~100

• With hit time finding, we can cope with 40-fold increase in luminosity

• For details see poster by Markus Friedl


Consequences for SVD Ladder Design

SVD2 ladders:• Up to 3 ganged (concatenated) sensors are read out from the side• Minimization of material budget, as hybrids are outside of acceptance• SNR > 15 with VA1TA, but would be < 10 with APV25

• Ganging of sensors does not work with APV25!

up to 3 gangedsensors

up to 3 gangedsensors


Motivation


Prototype Assembly


Summary & Outlook


Origami – Chip-on-Sensor Concept

(readout connections not shown)• Chip-on-sensor concept for double-sided readout

• Flex fan-out pieces wrapped to opposite side (hence “Origami“)• All chips aligned on one side single cooling pipe

Prototype for 4” DSSD(later with 6” sensors)


APV25(th inned to 100µm)

zylon ribcooling pipe

DSSDRohacellKapton

Origami – Details

side view

• Thinned readout chips (APV25) on sensor

• Strips of bottom side are connected by flex fanouts wrapped around the edge

• All readout chips are aligned single cooling pipe

• Shortest possible connections high signal-to-noise ratio

zylon rib

APV25 cooling pipe

3-layer kapton hybrid

integrated fanout(or: second metal)

DSSD

single-layer flex wrapped to p-side

Total material budget: 0.72% X0

(cf. 0.48% for conventional readout)


Belle-II SVD Layout

0

0

10

1+23

45

6[cm] layers

[cm]

20

-10-20-30 10 20 30 40

• Tentative geometry • New central pixel double-layer using DEPFET• 4 strip layers of 6” DSSDs• Every sensor is read out individually to maintain high SNR

– red: Origami chip-on-sensor concept– green: read out by conventional hybrid (from side)

Double-layer of DEPFET pixels

4 layers of double-sided strip sensors


Sketch of the Outermost Ladder (Layer 6)• Composed of 5 x 6” double-sided sensors

• Center sensors have Origami structure

• Edge sensors are conventionally read out from sides

Cooling block (end ring)

Connector (Nanonics)

Structural element (Zylon) Cooling pipe

Flex circuits

Electronics for border sensor

Electronics for border sensor

ca. 60cm


Motivation


Prototype Assembly


Summary & Outlook


APV25 Thinning

• One wafer (319 good dies) thinned down from 300 µm to 106 µm and diced by French company EDGETEK / WSI

• Received 314 good dies (5 lost = 98.4% yield)• 16 thinned APVs mounted onto 4 hybrids• Compared to 1 hybrid with 4 standard APVs

Measurement Results:• All APVs show similar signal and noise figure• Proper calibration curve• No measurable differences between normal

and thinned chips

Thinning has no effects on APV functionality and signal quality! APV25 Calibration Curve

Thinned APV25

Standard APV25


Origami Hybrid – Final Layout

4 n-side APV chips

2 p-side APV chips 2 p-side APV chips Connectors(on both sides)

Flex fanouts to beWrapped aroundthe sensor edge

• 3-layer flex hybrid design

• p- and n-sides are separated by 80V bias

• n-side pitch adapter is integrated in hybrid

• 3 pcs. of each type produced at CERN PCB workshop

Animal farm (mascots of creators)


Module Assembly 1Hamamatsu 4” DSSD:• top side: 152 µm pitch (z)• bottom side: 50 µm pitch (rφ)

• Gluing fanouts onto bottom side:– Custom jig (porous stone inlay) for

gluing and wire bonding– Two component epoxy paste

adhesive Araldite® 2011– Aligning fanouts against sensor and

hybrid– Distance between flexes is

important

• Wire bonding:– Fully automated bonder F&K

Delvotec 6400

• Processing of bottom side is finished


bottom APV25

top APV25

APV25 calibration curves of Origami hybrid

Module Assembly 2• Attaching APV25 chips onto hybrid

– Two component conductive adhesive

– Alignment under microscope

• Gluing hybrid onto Rohacell foam– There must be sufficient glue

underneath the bond pads!

• Wire bonding of APV25 power and control lines

• Electrical test of hybrid– some bad vias mostly repaired

with thin wires– Hybrid PCB design is okay– 7 of 8 APV chips are working well– Excellent internal calibration

results


Module Assembly 3

• Flip sensor and fanouts – Using another jig jig2– Jigs are stuck together with

3 alignment pins– Turning over– Remove jig1– Further assembling has to be

done on jig2• Gluing hybrid onto top side of

sensor– Aligning pitch adapter to strips

on sensor and fanouts• Wire bonding between sensor

and integrated pitch adaptor

2 openings to protect bottom side bond wires


Module Assembly 4

• Bending & Gluing of fanouts– Using positioner equipped

with a custom tool – Easier than expected– Allows precise positioning

w/o damaging underlying wire bonds

• Wire bonding between APVs and pitch adaptors

• Attaching cooling pipe– Thermal conductive paste– Electrically isolating pad

between APV and pipe


Final Origami Module

• Origami concept already presented at TWEPP 2008

• Prototype completed beginning of August 2009


Motivation


Prototype Assembly


Summary & Outlook


Beam Test Setup

• CERN SPS beam line Aug. 09• 120 GeV π+ / p / K+ • Cooling: 13 °C water

cooling system

beam

• Origami Module• Together with different types of

Belle DSSD prototype modules


Beam Test Results• Origami Module has been tested for several hours • Good performance w and w/o cooling (13 °C water)• About 80 GB data recorded• Analysis is in an early stage results preliminary• Benefit of cooling: ~ 10% higher SNR

120 GeV/c π+/p/K+ w/o cooling cooling (13° water)

p-side n-side p-side n-side

Average cluster width 2.2 1.8 2.2 1.8

Cluster SNR 11.5 16.7 12.8 18.5

PreliminaryPreliminary

• Cluster SNR = strip_signal / ( sqrt(cluster_width) * noise_avg)• ~10% higher noise compared to 2008 beam test for all modules

(ground loop in the front-end and / or HV supply?) • Hence SNR of all modules is slightly worse than expected


• Time resolution achieved in beam tests with several different types of Belle DSSD prototype modules (covering a broad range of SNR)

Beam Test Results 2

Time Resolution vs. Cluster SNR

0

1

2

3

4

5

6

7

5 10 15 20 25 30

Cluster SNR [1]

trm

s [n

s]Previous beam tests

SPS 09 beam test

Log-Log Fit

(TDC error subtracted)

Origami Module

• SPS 2009 data matches previous beam test.

• 2...3 ns RMS accuracy at typical cluster SNR(15...25)

• Origami shows similar precision like other modules


Motivation


Prototype Assembly


Summary & Outlook


• Motivated by KEKB upgrade and Belle II

• Concept for low-mass double-sided readout with cooling

• Thinning of APV25 chips does not affect its functionality

• Successfully built the first prototype of an Origami module

• Module showed excellent performance at beam test

• Beam test analysis is still ongoing

• Next goal: We want to build a prototype of the outermost layer of Belle-II SVD

Summary & Outlook


Thank you for your attention


BACKUP SLIDES


Origami Material Budget

• X0 comparison between conventional and chip-on-sensor:

Conventional (double layer kapton)

Layer Material X0 [mm] Thickness [mm] Percentage Area coverage Averaged PercentageSensor Silicon 93.7 0.3 0.32% 100.0% 0.320%Fanout Polyimide (2 layer of 50um each) 300.0 0.1 0.03% 96.3% 0.032%

Copper (10um) 14.0 0.01 0.07% 50.0% 0.036%Nickel (top: 1.3um) 14.3 0.0013 0.01% 50.0% 0.005%Gold (top: 0.8um) 3.4 0.0008 0.02% 50.0% 0.012%

Ribs Zylon (0.5mm wide) 300.0 5 1.67% 3.7% 0.062%Glue Araldite 2011 / Double sided tape 335.0 0.05 0.01% 96.3% 0.014%

Total 0.480%

DSSD Chip-on-Sensor (4-layer kapton)

Layer Material X0 [mm] Thickness [mm] Percentage Area coverage Averaged PercentageSensor Silicon 93.7 0.3 0.32% 100.0% 0.320%Isolation Rohacell (Degussa) 5450.0 1 0.02% 96.3% 0.018%Hybrid Polyimide (4 layers of 50um each) 300.0 0.2 0.07% 96.3% 0.064%

Copper (4 layers of 5um each) 14.0 0.02 0.14% 64.7% 0.092%Nickel (top: 1.3um) 14.3 0.0013 0.01% 64.7% 0.006%Flash Gold (top: 0.4um) 3.4 0.0004 0.01% 64.7% 0.008%

Flexes Polyimide (1 layer of 25um) 300.0 0.025 0.01% 56.3% 0.005%Copper (1 layer of 5um) 14.0 0.005 0.04% 28.1% 0.010%Nickel (top: 1.3um) 14.3 0.0013 0.01% 28.1% 0.003%Flash Gold (top: 0.4um) 3.4 0.0004 0.01% 28.1% 0.003%

8 * APV25 Silicon 93.7 0.1 0.11% 21.4% 0.023%SMDs SMD 50.0 0.4 0.80% 0.8% 0.007%Sil-Pad Sil-Pad 800 (Bergquist) 200.0 0.127 0.06% 11% 0.007%Pipe Aluminum (D=2.0mm, wall=0.2mm) 89.0 0.56 0.63% 7% 0.047%Rib Zylon (0.5mm wide) 300.0 5 1.67% 1.9% 0.031%Glue Araldite 2011 335.0 0.2 0.06% 50% 0.030%Cooling Water 360.5 1.26 0.35% 13% 0.047%

Total 0.719%

• +50% increase in material, but also huge improvement in SNR• Trade-off between material budget and SNR• According to simulation, additional material is prohibitive in

innermost layer, but no problem for layers 4-6 OK with layout


Origami Hybrid – Flexes

Documents

Construction and Performance of a Double-Sided Silicon Detector Module using the Origami Concept C. Irmler, M. Friedl, M. Pernicka HEPHY Vienna