Post on 13-Jan-2016
Columbia UniversityIN THE CITY OF NEW YORK
ROC3' / preproduction ROC Design Status
E.J. MannelStriPixel ReviewOctober 1, 2008
09/30/08
Eric J. Mannelmannel@nevis.columbia.edu
2
Columbia UniversityIN THE CITY OF NEW YORK
Strip Ladder Conceptual Design(Details to follow)
ROC ROC ROC ROC ROC BUS Ladder DataTransfer Board FEM
DCM
Top View of ladder
Bottom View of Ladder
RCC RCC RCC RCC RCC
Carbon Fiber Stave and Cooling(Light Blue) Optical Fiber
BUS
09/30/08
Eric J. Mannelmannel@nevis.columbia.edu
3
Columbia UniversityIN THE CITY OF NEW YORK
Strip Read Out Card (ROC)
Strip ROC consists of: StriPixel Silicon sensor
768 (6*128) X channels 768 (6*128) U channels split equally in to 3cm x 3cm region
12 SVX4 readout chips Passive component for SVX4 chips Interconnect with RCC module
09/30/08
Eric J. Mannelmannel@nevis.columbia.edu
4
Columbia UniversityIN THE CITY OF NEW YORK
interconnect
ROC Module-Concept
Stripixel Sensor
svx4
svx4
svx4
svx4
svx4
svx4
svx4
svx4
svx4
svx4
svx4
svx4
connector connector
Routing of SVX4Control/Data Signals
Location of Passive components
09/30/08
Eric J. Mannelmannel@nevis.columbia.edu
5
Columbia UniversityIN THE CITY OF NEW YORK
Readout Control Chip (RCC) Module Connects to the ROC- 1 RCC per ROC Contains Rad-Hard Readout Control Chip (RCC)-
Custom ASIC or FPGA Controls read out of the 12 SVX4 chips on ROC Multiplexes data onto readout bus Translates/passes LVDS signals to the SVX4s
Located on back side of strip ladder. Has passive components for power filtering. Connects to Ladder Data Transfer Board (LDTB)
via Ladder Bus Flex circuit construction
09/30/08
Eric J. Mannelmannel@nevis.columbia.edu
6
Columbia UniversityIN THE CITY OF NEW YORK
RCC Module-Concept
ROC Connector
ROC Connector
BusConnector
RCC
09/30/08
Eric J. Mannelmannel@nevis.columbia.edu
7
Columbia UniversityIN THE CITY OF NEW YORK
Ladder Bus
Connects multiple (5/6) RCC modules to Ladder Data transfer Board (LDTB)
Flex Circuit construction Mounted on bottom of ladder, extending
into Big-Wheel Region. No active components Possibly passive components for LVDS
termination (Details to be worked out)
09/30/08
Eric J. Mannelmannel@nevis.columbia.edu
8
Columbia UniversityIN THE CITY OF NEW YORK
Ladder Data Transfer Board(LDTB)
Located in Big-Wheel Region of VTX Rad-Tolerant FPGA
Translates the LVDS <-> Serial (SERDES) Strips channel numbers from the SVX4 data
Receives/transmits data from/to the FEM via optical fiber BCO-Mode Bits-Slow control data
Power regulation for ROC/RCC modules Rigid-Flex board design
09/30/08
Eric J. Mannelmannel@nevis.columbia.edu
9
Columbia UniversityIN THE CITY OF NEW YORK
Strip Front End Module (FEM)
Receives/Sends optical data from/to LDTB Data reduction and reformatting:
Real time pedestal subtraction Zero Suppression Adds PHENIX standard data headers and tails
Optical interface to DCM2 Optical interface with GTM
Sends BCO/Mode bits to LDTB Interface with slow control system
09/30/08
Eric J. Mannelmannel@nevis.columbia.edu
10
Columbia UniversityIN THE CITY OF NEW YORK
Strip Ladder Conceptual Design
ROC ROC ROC ROC ROC BUS Ladder DataTransfer Board FEM
DCM
Top View of ladder
Bottom View of Ladder
RCC RCC RCC RCC RCC
Carbon Fiber Stave and Cooling(Light Blue) Optical Fiber
BUS
09/30/08
Eric J. Mannelmannel@nevis.columbia.edu
11
Columbia UniversityIN THE CITY OF NEW YORK
ROC2
First attempt to place ReadOut Card (ROC) in top of stripixel sensor Observed pedestal shifts attributed to capacitive
coupling in the sensor and poor ground plane design
Internal review of the ROC2 in July 2007 Recommendations led to ROC3 design Rachid's presentation has shown results from the
ROC3 module
09/30/08
Eric J. Mannelmannel@nevis.columbia.edu
12
Columbia UniversityIN THE CITY OF NEW YORK
ROC3
Review Recommendations: Move readout chips to side of sensor Improve ground planes. Optimization of board stackup
Not intended as a ladder ROC First modules available early 2008 Used successfully in the FNAL Beam Test
(FNAL-T984) effort in the summer of 2008 Lessons learned were used for ROC3'
design
09/30/08
Eric J. Mannelmannel@nevis.columbia.edu
13
Columbia UniversityIN THE CITY OF NEW YORK
ROC3'
Minimal redesign of ROC3 to improve ease and yield of the fabrication and assembly of the ROC module
Verify reproducibility of ROC3 performance
Study the impact of changes in the overall board thickness and thickness of copper layers
Study the impact of moving/reducing passive components on the ROC
09/30/08
Eric J. Mannelmannel@nevis.columbia.edu
14
Columbia UniversityIN THE CITY OF NEW YORK
ROC3' Design
Maintain the same stack up as ROC31)Analog ground/ Components2)Analog Power/ Digital Ground3)Digital Signal4)Digital Power/Digital Signal5)Digital Ground
Wire bond SVX4 I/O pads directly to inner layer traces Requires a milled opening or cavity through first
two layers of board All components on top of the ROC
09/30/08
Eric J. Mannelmannel@nevis.columbia.edu
15
Columbia UniversityIN THE CITY OF NEW YORK
ROC3' Design Changes
Increased size of wire bond pads where possible
Relocated passive components to improve wire bonding if possible
Increase size of opening to access inner layer for wire bonding of SVX4 I/O connections
Reduce overall board thickness Produce board with both ½ oz and ¼ oz
copper layers
09/30/08
Eric J. Mannelmannel@nevis.columbia.edu
16
Columbia UniversityIN THE CITY OF NEW YORK
ROC3' Summary
ROC3' is a success: Analog performance matched or exceeded
ROC3 performance (Rachid's Talk) Easier to fabricate (Hughes Circuits) Easier to assemble (Hughes Circuits/BNL-
Instrumentation/SiDet) Thinner boards (overall/Cu thickness) perform
well Will reduce the overall radiation thickness of
the strip ladder
09/30/08
Eric J. Mannelmannel@nevis.columbia.edu
17
Columbia UniversityIN THE CITY OF NEW YORK
ROC3' Summary Still some fabrication and assembly issues
to be addressed: Attachment process of the bias plane to the ROC
still needs improvement. Opening to inner layers of ROC for wire bonding
of SVX4 still difficult. Clearances still an issue:
bond pad contamination (solder, flux, etc.) ¼ oz Cu version had yield problems.
ROC3' lessons lead to preproduction ROC design improvements
09/30/08
Eric J. Mannelmannel@nevis.columbia.edu
18
Columbia UniversityIN THE CITY OF NEW YORK
preproduction ROC (ppROC) Goals
Maintains or surpasses the ROC3/ROC3' performance.
Meets or exceeds the baseline design specifications: overall dimensions. radiation length.
Increasing the yield and reducing the cost. Not adversely effecting the VTX
installation schedule.
09/30/08
Eric J. Mannelmannel@nevis.columbia.edu
19
Columbia UniversityIN THE CITY OF NEW YORK
ppROC Design
Discussions with over the last month: VTX group to discuss global strip ladder
design Catalyst Microtech, Austin TX.
Wire Bonding Company Hughes Circuits, San Marcos CA
Board Manufacturer/Assembler HYTEC, Los Alamos NM
VTX/FVTX Engineering firm
09/30/08
Eric J. Mannelmannel@nevis.columbia.edu
20
Columbia UniversityIN THE CITY OF NEW YORK
Discussion Summary
Suggestions: Increase die pad size for SVX4 Solder, not glue, SVX4 to ROC Increase wire bond pad size Passive components on bottom, not top Increase keep out space around bond pads Eliminate wire bonding to inner layer of ROC Aluminum wire bonding is better then gold Larger clearances/traces for improved yield
09/30/08
Eric J. Mannelmannel@nevis.columbia.edu
21
Columbia UniversityIN THE CITY OF NEW YORK
Discussion Summary Cont....
Place ROC-RCC connection on bottom of ROC Design staves with no opening for bottom
mounted components Use stave design similar to ATLAS design Glue bus to bottom of stave Use some type of alignment pins to insure
alignment of ROC/RCC/Bus during assembly process.
Recommendations playing a leading role in design of the ppROC
09/30/08
Eric J. Mannelmannel@nevis.columbia.edu
22
Columbia UniversityIN THE CITY OF NEW YORK
ROC3' Passive Component Issue
High density of components between SVX4s:
Requires stacking of some components- 0201 on 0402
Interference with wire-bonding
Tight tolerance: 3 mil clearance
Chance that solder mask will not flow properly
09/30/08
Eric J. Mannelmannel@nevis.columbia.edu
23
Columbia UniversityIN THE CITY OF NEW YORK
ppROC Passive Component Solution
Move components to back side: More area available
for components No need to stack
components No issues with pad
contamination Allows repositioning
and enlargening of bond pads
Yellow top layerMagenta inner layersGreen bottom Layer
09/30/08
Eric J. Mannelmannel@nevis.columbia.edu
24
Columbia UniversityIN THE CITY OF NEW YORK
ROC3' Bias Plane Attachment Issue
Attached separately after surface mount components are attached by hand.
Required due to 0201 components on the top
Attachment technique results in uneven surface for svx4s and sensor
Labor intensive
09/30/08
Eric J. Mannelmannel@nevis.columbia.edu
25
Columbia UniversityIN THE CITY OF NEW YORK
ppROC Bias Plane Attachment Solution
Bias plane can be laminated on. Requires surface mount components on back side Allows use of solder balls in vias and re-flow to
make electrical connection between layers. Allows placement of connections for bias voltage
under the sensor Connection points should be level and not
interfere with sensor or SVX4s
09/30/08
Eric J. Mannelmannel@nevis.columbia.edu
26
Columbia UniversityIN THE CITY OF NEW YORK
ROC3' SVX4 I/O Pad Opening Issue
12 slots milled through top 2 layers to expose inner layer bond pads
Milled after board lamination.
Risk of damaging inner layer (yield)
Increases difficulty of wire-bonding.
Wire bond pads are narrow, 4 mils
09/30/08
Eric J. Mannelmannel@nevis.columbia.edu
27
Columbia UniversityIN THE CITY OF NEW YORK
ppROC SVX4 I/O Pad OpeningSolution
Move bond pads to top layer
Stagger pads to allow wider pads for bonding (not shown)
Reduces problems with wire bonding
Bond Pad regionTraces on top
09/30/08
Eric J. Mannelmannel@nevis.columbia.edu
28
Columbia UniversityIN THE CITY OF NEW YORK
ROC3' Overall Flatness Issue
Small amount of warpage in board Result of:
Thin boards Asymmetric stack-up
Very little can be done given the design-Inherent in the board design and fabrication process.
09/30/08
Eric J. Mannelmannel@nevis.columbia.edu
29
Columbia UniversityIN THE CITY OF NEW YORK
ROC3' ¼ Oz Copper Issues
Loss of electrical connectivity on board Result of fabrication steps
¼ oz Cu = .35 mil thick Different steps can remove some of the copper Cobra step most damaging Results in loss of electrical continuity on traces
09/30/08
Eric J. Mannelmannel@nevis.columbia.edu
30
Columbia UniversityIN THE CITY OF NEW YORK
ppROC ¼ Oz Copper Solution
½ oz Cu on trace planes. Wider traces where possible No need for Cobra step on trace planes.
part of the fabrication process Make ground and power planes ¼ oz Cu.
Will reduce overall board radiation thickness
09/30/08
Eric J. Mannelmannel@nevis.columbia.edu
31
Columbia UniversityIN THE CITY OF NEW YORK
ppROC-RCC Connection
Propose using interposer solution from Paricon Well established technology High reliability: Connection pads on RCC and ppROC are mated
with PariPoser material in between. Applied force maintains connection- requires
carbon fiber stiffener on ppROC and RCC Can be used for signals, power and bias
connections
09/30/08
Eric J. Mannelmannel@nevis.columbia.edu
32
Columbia UniversityIN THE CITY OF NEW YORK
ppROC-RCC Connection
Silicone insulator with vertical conduction paths
Placed between pads, compression makes the connection between surfaces
Alignment only of circuit boards required
09/30/08
Eric J. Mannelmannel@nevis.columbia.edu
33
Columbia UniversityIN THE CITY OF NEW YORK
ppROC-RCC Interconnection
09/30/08
Eric J. Mannelmannel@nevis.columbia.edu
34
Columbia UniversityIN THE CITY OF NEW YORK
Mechanical
Board Dimensions: Wide board design is baseline (SVX4s on the side
of the sensor). Maintain 80mm board width base line. Maintain 65mm board length base line
Baseline dimensions can be met: ppROC-RCC connection grid on bottom of ROC No flex cable to wrap around stave Component placement is still tight
09/30/08
Eric J. Mannelmannel@nevis.columbia.edu
35
Columbia UniversityIN THE CITY OF NEW YORK
Mechanical Dimensions
09/30/08
Eric J. Mannelmannel@nevis.columbia.edu
36
Columbia UniversityIN THE CITY OF NEW YORK
ppROC Work in Progress
New Schematic almost complete Changes for final ppROC-RCC connection required Move power filtering caps to the RCC
ppROC layout has started, but still work to be done. New ppROC-RCC connector Changes to SVX4 die pad size Change location and size of bond pads Change bias plane connections
09/30/08
Eric J. Mannelmannel@nevis.columbia.edu
38
Columbia UniversityIN THE CITY OF NEW YORK
ppROC Schedule
Layout complete in early 4Q CY-2008 First ppROCs fabricated late 4Q CY-2008 First strip modules assembled early 1Q
CY-2009 First strip modules ready for system chain
test late 1Q CY-2009
09/30/08
Eric J. Mannelmannel@nevis.columbia.edu
39
Columbia UniversityIN THE CITY OF NEW YORK
Conclusions
ROC3' performed as well as ROC3 Lessons learned from ROC3' will improve
ppROC design: Ease of fabrication and assembly Result in improved yield and lower cost
ppROC design will maintain baseline ladder design constaints and meet performance requirements
Time line is consistent with chain test in 1Q CY 2009
09/30/08
Eric J. Mannelmannel@nevis.columbia.edu
40
Columbia UniversityIN THE CITY OF NEW YORK
Backup
09/30/08
Eric J. Mannelmannel@nevis.columbia.edu
41
Columbia UniversityIN THE CITY OF NEW YORK
Concerns/Issues
Test of removing “extra” filter caps showed widening of pedestals for some SVX4s