lbnl design studies

of 49 /49
M. Gilchriese LBNL Design Studies M. Garcia-Sciveres and M. Gilchriese LBNL W. Miller and W. Miller, Jr iTi

Author: veta

Post on 12-Jan-2016




0 download

Embed Size (px)


LBNL Design Studies. M. Garcia-Sciveres and M. Gilchriese LBNL W. Miller and W. Miller, Jr iTi. Introduction. Layout studies along with thermal and mechanical modeling (FEA) were started some months ago. - PowerPoint PPT Presentation


Page 1: LBNL Design Studies

M. Gilchriese

LBNL Design Studies

M. Garcia-Sciveres and M. GilchrieseLBNL

W. Miller and W. Miller, JriTi

Page 2: LBNL Design Studies

M. Gilchriese2


• Layout studies along with thermal and mechanical modeling (FEA) were started some months ago.

• Primarily aimed at defining R&D direction(s) for new materials for module support/cooling structures.

• Only in the last few weeks have we tried to study cases directly applicable to – Case A/1(inside current B-layer)

– Case B/2(remove and replace B-layer)

• Will only show single-layer studies today, although some work has been done on double-layer designs.

• Please keep in mind that assumptions about module sizes and other dimensions have evolved during these studies.

Page 3: LBNL Design Studies

M. Gilchriese3

Overview – Design Studies

• Integrated monolithic structure– Many modules mounted on half-shell units ie. two pieces make a B-layer

• Staves supported on shells. Similar concept to current detector but different implementation.

• All options studied in some detail assume for module supports/cooling– Round tube embedded in low-density conducting foam(requires some

development – see end of talk) + very thin “skins” of carbon-fiber laminates(eg. K13D2U)

– Why round tube? Would allow full-pressure testing with gas(not possible in current detector).

– Why foam? Versatile. Low-mass. Compatible with both monolithic and stave-based designs.

• Usually we have assumed a heat flux from module(not including cables) of 0.6 W/cm2. Current estimates of electronics+sensor are lower than this.

Page 4: LBNL Design Studies

M. Gilchriese4

Integrated Monolithic Option

• Support split into two halves

• 35 mm radius inner stay clear

• Low density foam structure with thin(125 micron) skins on inside and outside(makes shell).

• Heat load assumption for 800mm length=120W

• Two pass for each cooling tube: 5mm ID to limit pressure drop to <200mbar 88mm



Page 5: LBNL Design Studies

M. Gilchriese5

FEA Studies• See backup for most details and figures

• Gravity sag over 800 mm length: about 3 microns

• Distortion from cooldown(T=50C): <10 microns

• Thermal performance– Tube wall -22C

– CGL7018 coupling tube to foam

– Detector temperature -18 to -15

– Note red is IC “overhang”

• Proof-of-concept



Center detector more representative

Page 6: LBNL Design Studies

M. Gilchriese6

Staves With Support Shell• Confined space, need additional room for support shell

– Provide stave with 5 point support

– Minimize amount of construction material• Combination of high conductivity foam as before in the integrated

design. And two layer laminate, uni-tape or single layer of woven cloth

• What happens to interfacial stresses(tube to foam)– Calculated, but best resolved through testing

– Have not looked at shell in any significant detail yet

• Module concept and dimensions evolved over time as we studied this direction – see next slides for assumptions and also more in backup slides.

Page 7: LBNL Design Studies

M. Gilchriese7

Bare Module Dimensions


Active area










200um chip

20um bumps

250um sensor

Module power load: 1.25W => 0.4 W/cm2 but we use 0.6 W/cm2

Page 8: LBNL Design Studies

M. Gilchriese8

Stave Arrangement

Modules abut one against the next.

Flat stave surface!

48 modules per stave for 778mm total length. Readout/power from both ends.

Page 9: LBNL Design Studies

M. Gilchriese9

Option 1: Module with Flex Cable14



Flex cable



Cable power load: 0.4W (uniform over full 50cm)

Each module has a pre-attached, full length cable All cables are identical

Page 10: LBNL Design Studies

M. Gilchriese10

Option 2: Cable(s) pre-laminated on back of stave

Flex would be laminated to back side of stave and tested before modules are added A flap at each module location is bent around and to the top of the module and wire bonded after the module is loaded.

A single 2-sided copper cable could be used to route all the signals. Multiple aluminum on kapton planes could be laid up on top of that to build up power bussing, bonding each one down to the copper flex to get the power to each flap. Recall all this is built and tested before modules are added, so repeated gluing and wire bonding are not an issue. Assume same cable power load as for option 1. A stave control card would be loaded at the stave end as for option 1

Signal cable with 24 flaps before lamination. Could also be multiple cables.

End of stave card could be built-in

Page 11: LBNL Design Studies

M. Gilchriese11

Initial Layout-Stave 1

• Concept– Retained features of integrated

design, same cooling tube size

– Less foam, but added cylinder

– Outer diameter ~93mm

– Inner diameter 70mm

Page 12: LBNL Design Studies

M. Gilchriese12

Stave Concept 1 for B-Layer Replacement

Half-length shown

Page 13: LBNL Design Studies

M. Gilchriese13

Stave 1: Basic FEA Configuration

• Effects simulated– Mass of coolant, average density 145kg/m3

– Laminate, 2 layers 2.5mil, 0/90, K13D2U

• Radiation Length estimate=0.532%– Foam=0.11%

– Tube=0.3%

– Composite=0.11%

– Coolant=0.012%5mil laminate

5.6mm OD tube12mil wall

0.5mm of silicon to simulate chips and detectorAlso 0.9 mm of Kapton cable for additional mass

Page 14: LBNL Design Studies

M. Gilchriese14

Second Configuration-Stave 2• Goals

– Reduce tube size and amount of foam material

– Analytically evaluate impact on thermal and mechanical design

OD=88 mmID=~69.75mm

Page 15: LBNL Design Studies

M. Gilchriese15

Stave 2 Concept

Page 16: LBNL Design Studies

M. Gilchriese16

Stave 2: With Offset Mounts• Space on back-side next to mounts appears adequate to place

cable, for wrap-around mounting

Cable position(for thermal analysis)Potential cable


Page 17: LBNL Design Studies

M. Gilchriese17

Stave 2: With Offset Mounts• Cable Illustration

– Back side, thin bonding wraps around

– Cable on back-side becomes thicker as the stave end is approached

Cable constant thickness in this region

Page 18: LBNL Design Studies

M. Gilchriese18

Stave Concepts - FEA Summary

Stave 1 Stave 2 Comments

Gravity Sag Negligible <0.1µ Negligible <0.5µ Does not include shell sag! 5-pt support

Cooldown distortion

<6 µ out of plane About 50 µ 50C change

Stress in foam-tube interface

300 psi 145 psi 50C change. Does not include glue compliance

T(C)-no cable About 5 About 5 Tube at -22C

T(C)-with cable on front

About 6 About 6 Worst case?

Page 19: LBNL Design Studies

M. Gilchriese19

“Powerpoint” Designs

• Castellated monolithic– Not analyzed yet

• Rectangular tube – gain about 1mm on outside radius(OR) – Stave 2 change OR 44->43

• Rectangular tube with monolithic maybe gains 1 more mm on OR to 42mm

• Minimal monolithic design(imagine from figure below) might get to R=42 mm.

R=41.9 to back

Page 20: LBNL Design Studies

M. Gilchriese20

Comparison with Cases A/BInner Radius(mm) Outer Radius(mm)

Case A now 35 41

Case A future 27 41

Case B now 35 71.5

Case B future 27 71.5

1st monolithic 35 44

Stave 1 35 46.5

Stave 2 35 44

2nd monolithic,rect. tube….

35 42? TBD

Page 21: LBNL Design Studies

M. Gilchriese21

Development Direction• Carbon foams with good thermal conductivity, but significant

density, are available from multiple producersPOCO foam: eg. = 0.55 g/cc and K(out) 135 and K(in) 45K-foam: eg. = 0.34 g/cc and K(out) 55 and K(in) ?

• We have, in fact, made staves with POCO foam/round tube as part of upgrade R&D for outer silicon tracker.

• We are working with company to make samples with about 0.15 g/cc and K of about 45(isotropic). In production now.

• Would make short prototype staves and look at thermal performance.

• In addition, are hoping to also investigate carbon nanotube(CNT) loaded materials with same company and perhaps CNT “cloth” under development for heat spreaders for ICs. CNT have very high K along tube direction >1000.

Page 22: LBNL Design Studies

M. Gilchriese22

Conclusion• Integrated monolithic concept appears to be structurally and

thermally feasible• Multiple stave concepts developed and also feasible(no

surprise)• All based on edge-to-edge modules (no shingling). Need to

confirm this should be design choice. • Looks challenging to meet Case A(current constraints)

envelopes. Need iteration of envelopes, design, maybe beam pipe, as-built dimensions….before investing in detailed design of any option.

• Low-density, thermally conductive foam with very thin carbon-fiber facings appears to be feasible approach mechanically and thermally.

• Prototypes to be made to validate approach.

Page 23: LBNL Design Studies

M. Gilchriese


Page 24: LBNL Design Studies

M. Gilchriese24

Integrated Monolithic Structure

Page 25: LBNL Design Studies

M. Gilchriese25

Integrated Structure Assumptions• Split Structure

– Sandwich structure, with cooling tubes embedded between 2-layer composite facing

• Composite laminate produced using K13D2U fibers and Cyanate Ester resin

– 5mils for two layers (0/90)

• 5 mm ID Aluminum tubing, 12 mil wall (~5.6 mm OD)

• FEA Structural Model– Tubes and foam core treated as solid elements

• Mass of coolant, average density 145kg/m3

– Outer surface laminate: used laminate element, with single material

– Inner surface (saw-tooth) contain laminate elements with material designations for:

• Composite layers (0/90)

• Silicon module assembly, 0.5mm silicon

• Cable, 0.9mm uniform along length

Page 26: LBNL Design Studies

M. Gilchriese26

Gravity Sag• Model based on 1G loading vertical

– Sag measured in local coordinates– T1: translation is vertical along shell split plane

– Maximum sag ~2.8microns

– Model length 800mm


Page 27: LBNL Design Studies

M. Gilchriese27

Thermal: 50C Temperature Change


• Thermal strain due to cool-down– Local coordinates, T2 is transverse to

vertical plane of symmetry• peak shape change is 5.5microns

– Model length 800mm

Unfortunately the out-plane distortion is a combination of T1 an T2

Page 28: LBNL Design Studies

M. Gilchriese28

Thermal: 50C Temperature Change


• Thermal strain due to cool-down– X: direction 8.2 to 6.6 microns

• X is split plane, using symmetry boundary conditions

– Model length 800mm

Page 29: LBNL Design Studies

M. Gilchriese29

Pixel Thermal Solution-Integrated Structure

• Description– Isotropic carbon foam: 45W/mK

• Specialized low density (0.21g/cc) foam: enhanced to high conductivity

– Includes 5mil laminate thickness– Detector 250microns– Chips 200 microns– Bump bonds 25microns– Interface resistance from bonding chip

to foam equal to 0.8W/mK; 4mil thickness (CGL7018)

– Pixel chip heating: 0.51W/cm2

– Simulated tube wall -22ºC

• Results– Peak chip edge: -13.8 ºC– Detector ranges from-15.6 to -17.8 ºC



Center detector more representative

Page 30: LBNL Design Studies

M. Gilchriese

Module design for B-layer replacement


Page 31: LBNL Design Studies

M. Gilchriese31

Option 1 module loading

• For even number modules the cable is cut short, to reach only its odd neighbor to the left (see sketch below)• For odd modules the cable is cut long enough to reach the end of the stave• All cables are wire-bonded to a card at the end of the stave. • Each cable reaching the end of the stave serves 2 modules• For proper cable stacking, loading starts from the outside and moves to the center

4 62 3 51End of stave card

Section through modules and cables

Page 32: LBNL Design Studies

M. Gilchriese32

Option 1 cable power

• Power load of 0.4W/module estimated assuming 0.5V R/T drop in cable at 800mA total current.

• If all cables are identical, this means 8mW/cm of cable per module. • Cable load at central module is 8mW/cm x 1.62cm = 13mW• Cable load at end module is 13mW x 24 modules = 312mW. • Cable load at Nth module counting from stave center is 13N mW. • Reduction of material at stave ends could be achieved by ganging more modules together on 1

cable. • Doubling modules/cable doubles cable load/cm/module. If this is done for last 12 modules,

cable power at end module increases to 12x13 + 12x2x13 = 468mW.

Page 33: LBNL Design Studies

M. Gilchriese33

Stave control card & external cables

• Assume each 1-chip module receives 40MHz clock and command, and outputs 160MHz data.

• The end of stave card reduces the number of lines to the outside world by serializing the data from multiple modules.

• Some or all DC-DC converters could be placed on end-of stave card (=> end of stave card needs good cooling potentially up to ~10W).







40MHz clock 40MHz clock bar

40MHz command

40MHz command bar

160MHz data

160MHz data bar

Clock fanout

End of stave cardTraces on flex~15m miniature

coax to PP2 40MHz clock

1GHz command

2GHz data

2GHz data

LV power

command De-serializer

2 Data serializers

Power at “HV”12-24DC-DC

4-12HV bias groups 4-12HV bias groups

3Muxed NTCs

Twisted pairs

NTC mux

Page 34: LBNL Design Studies

M. Gilchriese34

3-wire hardware interlock compatible NTC mux scheme









Page 35: LBNL Design Studies

M. Gilchriese35

Stave Approach

Page 36: LBNL Design Studies

M. Gilchriese36

Stave 1: Gravity Sag

• Upper Stave position near vertical centerline– Modeled ½ length, from mid

plane of symmetry of a 778 mm long stave

– Model provides effect of a 5 point support stave length

• Resulting gravity sag 0.085microns

Does not include support shell sag

Page 37: LBNL Design Studies

M. Gilchriese37

Stave 1: Thermal Distortion

• Cool-down effect: 50°C Delta– Most of distortion is

contraction along stave length– Distortion T2 is out-of-plane

– T2 peak distortion is 5.6microns

Page 38: LBNL Design Studies

M. Gilchriese38

Stave 1: Thermal Strain• Stress in Foam/Tube Interface

– Evaluated without compliance of bonding adhesive (CGL7018 type)

– Contraction of Al tube produces local stress of 300psi at interface • Effect best evaluated through testing

– Plan is to use special Reticulated Vitreous Carbon Foam with enhanced thermal and mechanical properties

Solid Von Mises Stress

Page 39: LBNL Design Studies

M. Gilchriese39

Stave 2: Gravity Sag• Upper Stave position near vertical centerline

– Modeled ½ length, from mid plane of symmetry of a 778 mm long stave

– Model provides effect of a 5 point support stave length

• Resulting gravity sag 0.41microns

Does not include support shell sag

Page 40: LBNL Design Studies

M. Gilchriese40

Stave 2: Gravity Sag-Off Set Mount

• Effect on rotation of stave– Maximum rotation 1.9μradians due to sag

Page 41: LBNL Design Studies

M. Gilchriese41

Stave 2: Thermal Distortion• Stave with out-of-plane bending due to cool-down 50°C

– Modeled ½ length, from mid plane of symmetry of a 778 mm long stave

– Model provides effect of a 5 point support stave length

• Resulting bending 51.5 microns

Page 42: LBNL Design Studies

M. Gilchriese42

Stave 2: Thermal Strain• Stress induced by contraction

– Less than in Stave 1 geometry– 145psi, more localized at ends– Be mindful that compliance of adhesive not present

Page 43: LBNL Design Studies

M. Gilchriese43

Stave 2: Thermal Solution

• Model Parameters– Carbon Foam, 45W/mK

– Composite Facing, K13D2U-55% vol fraction• 0/90, Kt=0.55W/mk, 220W/mK planar (no axial thermal gradient so

this parameter is not an issue)

– Chip 0.2mm

– Bump bond thickness, .05mm

– Detector, 0.25mm

• Adhesives– Tube to foam, 4mils, 0.8W/mK

– Foam to composite facing, 2mils, 0.8W/mK

– Chip to composite facing, 4mil, 1.29W/mK

– Cable to detector module, 2mils, 1.55 W/mK

Page 44: LBNL Design Studies

M. Gilchriese44

Stave 2: Thermal Solution-No Cable• Coolant Tube “BC”

– -22ºC

• Chip Heat Flux, 0.6W/cm2

• Detector Temperatures– Left edge, -16.66ºC

– Middle, -17.01ºC

– Right, -16.78ºC

Page 45: LBNL Design Studies

M. Gilchriese45

Stave 2: Thermal Solution-With Cable• Cable heat load

– Adds a heat flux of 0.1W/cm2 to the 0.6W/cm2 chip heat load

– Gradient before was 4.99ºC, detector middle to tube inner surface

– Would expect gradient of 5.82ºC now

– Gradient now from detector middle to tube surface is 6.0ºC

• Cable surface– Peak -14.1ºC, or a ΔT=7.9ºC– Peak affected by K assumed

for the copper/Kapton cable• Used 0.35W/mK, whereas

Kapton alone is 0.12

Page 46: LBNL Design Studies

M. Gilchriese46

Stave 2: Thermal Solution-With Cable

• Thermal plot with cable removed– Illustrates comparative

uniformity in detector temperature



Page 47: LBNL Design Studies

M. Gilchriese47

Detector Temperature Summary• Thermal Solutions for two designs, but unfortunately different

detector layouts– Integrated, different by chip over-hang

– Stave-like, provides complete coverage• Two different foam/sandwich structures, one with less material analyzed


• With time will bring configurations into consistency• However, the predicted detector surface temperature for each is:

– Low-mass stave without cable heat load, -17ºC

– Low-mass stave with cable heat load, -16ºC

– Integrated Foam/Tube Support without cable load, -17.5ºC• Caution, as analysis proceeded slightly more conservative properties were

used for the composite facing and the foam:– Facing 0.55W/mK versus 1.44W/mK

– Foam 45W/mK versus 50W/mK

Page 48: LBNL Design Studies

M. Gilchriese48

Suggestions for Future Work• Benefits of Continued Analysis

– Improve definition of concept• Preliminary results encouraging, but issues will emerge with increased

knowledge base

• Add support shell deflection, support boundary conditions, end plates etc.

– Cooling analysis• Conservatively used 0.6W/cm2 for chip thermal load, revise as electronics

design progresses, also expand on cable thermal load (location)

• Tube sizing analysis needs more careful consideration– Preliminary pressure drop with smaller tube appears OK with C3F8

» An area that needs more study, pressure drop prediction indicates low margin on return pressure

– Consider tube shape change to reduce outer diameter (<88mm?)

• Prototyping– Thermal and structural prototypes

Page 49: LBNL Design Studies

M. Gilchriese49

Suggestions for Future Work (Cont.)

• Basic Stave Concept uses Lightweight Sandwich– Continued Material development Important to Success

• Carbon foam development– Have produced 45W/mK foam

» By very nature of HEP goals, heavier than desired

• Incentive: Advance through development with carbon nano-tubes additives– Stronger, lighter, more conductive?