Hardware Lessons Learned from Building Beagle

1
Hardware Lessons Learned from Building Beagle
Gerald Coley ESC-401
2
Topics Covered

• Whats in a name?
• Specification
• Schematic
• PCB Layout
• Assembly
• Testing
• Support
• Implementing Lessons Learned
• Questions and discussion

3
The NameWhere did we get the nameBeagleBoard?
4
How did we arrive at BeagleBoard?

• Every project needs a name
• Named for my dog.
• Just temporary
• We will fix it later
• Well, It Stuck
• A Beagle is
• Like Tux, non threatening
• Curious
• Loyal
• Fun
• And gives back as much as it receives
• Lesson LearnedIf you name it, it will stick!

5
So, we need an acronym

• Bring your own peripherals
• Entry-level
• ARM Cortex-A8
• Graphics DSP
• Linux and open source
• Environment for SW innovators

6
The SpecificationA buffet or just meat and
potatoes?
7
The Goals

• Get OMAP3530 into peoples hands
• Low Cost lt150
• Under the spousal radar
• Reach as many people as we can
• Small Size
• Keeps the cost down
• Cool Factor
• Internally focused
• Cortex-A8 NEON
• DSP
• 3D Graphics
• Accelerators

8
First Pass Features

• OMAP3530 Processor
• 3.5 LCD
• VGA/QVGA
• Touch screen
• S-Video
• Keypad
• 2 Serial Ports
• SD/MMC
• Ethernet
• Camera
• 128MB DDR
• 128MB NAND
• USB
• Host
• OTG
• USB/DC Power
• Stereo In/Out
• Battery w/Charger
• Full Expansion Bus

COST 1500 Size 5 x 8
Something has to go!!
9
Bring Your Own

• Not everyone needs everything
• Everything is needed
• Bring your Own
• User adds only what they need
• They dont have to pay for what they already have
• Allows for maximum exposure to HW
• More than just one component supplier

10
Final Pass Features

• OMAP3530?
• 3.5 LCD.-VGA/QVGA?
• Touch screen USB Touch Screen ?
• DVI-D VGA, SVGA, XGA ?
• S-Video ?
• Keypad
• 1 button USB Keyboard Mouse ?
• 2 Serial Ports
• 1 Port USB Expansion ?
• SD/MMC 6 in 1 Slot ?
• Ethernet
• USB Dongle WiFI ?
• Camera
• USB Camera ?
• 128MB DDR ?
• 128MB NAND ?
• USB Host ?OTG?
• USB/DC Power ?
• Stereo In/Out ?

Standard on Beagle Bring your own
COST 149
? Just meat and potatoes, but still room for
dessert!!
11
The Schematic
12
Play It Safe

• Leverage everything we can
• Used OMAP35xx EVM as base
• Maintain compatibility with other platforms
• Common hookup
• Subset used the same
• Keep the component count low
• Use what is proven to work
• Leave off what is not needed

13
Leveraged Advantages

• OMAP35xx Processor Symbol/Design verified
• TWL4030/TPS65950 Symbol/Design verified
• Removed a lot of features here
• SD/MMC Verified
• DVI-D verified
• UART Verified
• Lowered the overall risk

14
PCB Layout
15
Rules We Used
PCB Design
PCB FAB
PCB ASSEMBLY
TI
Memory Vendors
Failure to include the fab and assembly team
members can prove costly since choices made early
in the design will adversely impact the final
cost of the assembled board
16
The Challenges

• OMAP35xx Package
• .4mm Pitch
• Routing
• TPS65950(TWL4030)
• .4mm pitch PMIC
• POP Implications
• Limited area for connectors
• Location driven and not layout driven

17
OMAP3530 Package
18
What Exactly Is POP?
Memory
OMAP
19
PMIC TPS65950 (TWL4030)
20
Fixed Location of Connectors

• Routing complicated by locations
• Connector not always on the best side
• We tried to make this as easy as possible
• Not the best locations for routing
• The DVI-D was the longest route
• We need to keep the layer count down
• Our goal was to stick with 6 layers

21
Recommended Trace and Via Connections
Poor
Better
Watch for solder traps
22
Blind and Buried Vias

• Definitions and Terminology
• Through hole via has access to both external
  layers
• Buried via provides connection within inner
  layers
• Blind via does not pass through the entire board

23
Via-in-Pad
Places the via directly in the BGA pad Greatly
improves board routing Has special requirements
for manufacturability

• Given a 10mil BGA pad
• Use a 4mil microvia
• Laser drilled
• Via must be filled or capped to minimize void
  formation
• Via-fill Material
• Check with board fabricator for their
  preferences
• Fill material can be conductive or
  non-conductive
• Via fill materials particle size must be lt1mil
  for complete penetration

You must validate your board fabricators
capability to reliably build with this class of
technology
24
.4mm Pitch

• No routing between pads
• Use top layer routing on outside pads
• Use vias-in-pad wherever needed

25
Soldermask

• Exposed pad should be the same size as the pad on
  the mounted device
• Need to make the pad larger to add stability
• Calls for soldermask defined pads

26
Surface Finish

• The PCB surface finish provides a coating over
  the outer layer copper that prevents oxidation
  and provides an electrically conductive surface.
• Organic Solderability Preservative (OSP) Thin
  layer of organic material to prevent copper
  oxidation. It is removed by the assembly flux.
  Does not tolerate multiple heat cycles.
• Immersion Tin (ImSn) Thin layer of tin directly
  on top of the copper surface. Produces an
  extremely flat surface for mounting of surface
  mount components. Downside is the possible
  formation of Tin whiskers.
• Immersion Silver (ImAg) Thin layer of silver
  directly on top of the copper surface. Produces a
  very flat surface. Compatible with no-clean
  assembly processes. Maintains high solderability
  after multiple heat cycles. Downside is that the
  plating will tarnish over time.
• Electroless Nickel Immersion Gold (ENIG) Commonly
  used finish that is nice and flat for fine-pitch
  devices. Not easily reworked and more expensive.
• Hot-air solder leveling (HASL) Immersing the PCB
  in solder. It is inexpensive and widely
  available, but it is not flat and therefore does
  not work well with fine-pitch devices

27
Final BeagleBoard Design
28
OMAP3530 BGA Footprint
29
Via-In-Pad Topology

• Stacked Microvia-In-Pad
• Pad 11mils copper (10mil SMD)
• Top-Layer2 stacked with Layer2-Layer-3
• Top-Layer2
• Laser Drill 6mil (0.152mm)
• Solder-mask defined pads
• Non-conductive epoxy via plugs, plated over and
  planarized

BeagleBoard does not use buried vias
OMAP3
Top Soldermask
Top Signal
Blind 1-2
Blind/Stacked 1-3
Lyr2-GND
Lyr3-SIG
Lyr4-SIG
Lyr5-PWR
Bottom Signal
Bottom Soldermask
30
Standard Vias

• Thru-hole via from top to bottom
• 18mil radius pad
• 8 mil drilled hole
• Plugged, plated and planarized
• Must be level
• No dimples

Top
Lyr2-GND
Lyr3-SIG
Lyr4-SIG
Lyr5-PWR
Bottom
Not To Scale
31
Top Layer-Signal (Area under OMAP3 Chip)
Via 1-6
V-I-P
BGA PAD
3mil trace
10mil trace
32
Bottom Layer Bypass Capacitor Locations
Capacitor Spec Type Ceramic X7R Size 402 Value
0.1uF
33
PCB Fabrication Concerns

• High Board cost
• Because they can charge more
• Unknown
• Does not necessarily mean the production will
  be high
• Unfamiliar with fine pitch
• Soldermask registration critical

34
Board AssemblyPutting it all together
35
POP Assembly Concerns

• POP was the number one concern
• Nitrogen and air were the two options
• What should we do?
• Picked the one that was the most applicable to as
  many assembly houses as possible.Air
• POP soldering in air
• Air 1
• Air 2

36
Mounting .4mm Parts

• No real concerns
• Equipment can handle it
• Warping of the processor due to POP was a
  possibility
• Saw no reason to be concerned
• There could be an issue with shorts if soldermask
  not correct
• We were comfortable

37
Solderpaste Equipment
38
Pick and Place Machine
My DATA (MY9) D-014-1486 F30
39
Reflow Oven
Heller EXL
40
X-Ray Machine
Glenbrook Technologies RTX-113
41
POP Mounting Configurations

• Method A
• POP onto OMAP3 first? reflow
• Back side Second? reflow
• Top side third w/POPOMAP3530? reflow
• Method B
• Mount back side ? reflow
• Mount top side with OMAP3530 ? reflow
• Mount Memory w/ Pick Place ? reflow
• Method C
• Mount backside ? reflow
• Mount topside with OMAP3530 and POP ? reflow

42
Method A Process
Hand Mount
TopGEL
POP
Mount w/SMT
OMAP3
OMAP3
Paste
SMT
Reflow
Reflow
Paste
Top side of board
Bottom side of board
Hand Assembly
43
Method B Process
Paste
Paste
SMT
Reflow
SMT
Reflow
Top side of board
Bottom side of board
ApplyGel
POP Memory
Reflow
SMT
Hand Assembly
44
Method C Process
Dipping Arm is used to apply paste
POP Memory
POP Memory
Paste
Paste
SMT
Reflow
Reflow
SMT
Top side of board
Bottom side of board
Hand Assembly
45
POP Mounting Analysis

• Method A
• No real issues seen
• We abandoned it when we ran into issues
• Issues were not related to this process
• Method B
• Final method adopted
• Have had excellent results
• Method C
• Dipping ArmPOP Memory dipped into GEL
• Have a lot of high volume customers doing this
• Did not try as we had no dipping arm

46
Panelization

• Method A
• 4 boards per panel
• Used it on the first runs
• Had concerns about warping across the scoring
  that could cause solder shorts
• Nothing ever proven that this was an issue
• Method B
• One board per panel
• Method chosen

47
Other Parameters

• Stencil thickness
• .4mils
• Solder Paste
• AMTECH LF -4300 Lead Free
• Chemistry Sn96.5/Ag3.0/Cu.5
• Tacky paste flux (memory)
• Amtech Tacky Solder Flux

48
Reflow Profile Diagram
Time
49
Board Profile Settings
50
Issues on 1st Run

• Solder Shorts under the processor
• Always a power rail
• 10 Yields
• Could it be related to.?
• POP Package
• Profile
• PCB
• Parts
• Solder

51
Solder Shorts
52
2nd Run

• What is the easiest thing to change?
• Adjusted the temperature profile
• Same results
• 90 Fail
• Shorts under processor

53
3rd Run

• Shotgun approach
• Changed PCB Vendor
• Went to single board penalization
• Changed POP method from A to B
• Kept the original profile
• Results
• 96 Yields
• No Shorts

54
Analysis of 2nd Run

• What was the issue?
• Gut said PCB soldermask
• Contacted first supplier
• No issue w/soldermask per supplier
• Ordered more board and ran again with all other
  methods from Run 3
• Same problem 90 fail
• Ordered shotguns and raided first supplier
• They admitted that they opened the soldermask
• Had them run another batch w/o changing
  soldermask
• Result
• 96 yields
• No shorts

55
Good/Bad Soldermask
GOOD
BAD
56
Other Issues Along the Way

• SMT PCB audio connectors can rip off
• Only happened on two boards
• Pads detached from the board
• Added clear epoxy to connectors
• Adding vias to pads to add strength (Rev B6)
• SMT USB connectors can rip off
• Only happened on one board
• Adding vias to pads to add strength (Rev B6)

57
Final Analysis

• Soldermask is the key issue
• PCB suppliers do their own thing
• Make sure that the thing they do is your thing!
• All three POP methods should work
• Panalization should not be an issue, but not
  confirmed
• It will be an issue if the soldermask is bad
• Beagle profile is a good place to start
• Will need to be adjusted based on equipment used

58
Board Testing Process
59
Inspection Points

• PCB
• Visual inspection
• Solder Paste
• Bottom side application
• Top side application
• Shorts
• After re-flow
• AOI
• After final assembly
• Manual inspection
• Final inspection

60
Assembly Inspection Points
Paste
PCB
Paste
Paste
Shorts
Paste
SMT
Reflow
SMT
Reflow
Top side of board
Bottom side of board
ApplyGel
POP Memory
Visual
Shorts
AOI
Reflow
SMT
Hand Assembly
61
PCB Check

• Check soldermask
• Overlap (Soldermask defined pads)
• Soldermask over vias
• Check finish
• Discoloring of finish
• Dimples in via in pads
• Must be smooth
• Check Pads
• Look for dimples in the vias in pad areas

62
Solderpaste Inspection
Check paste for smooth application Check for
missing paste (not sticking to PCB)
63
Shorts Test

• Implemented when we had solder shorts
• Easy check for main issues we have seen
• Check across the caps using a ohmmeter
• Run as a spot check
• Can be done at 100

64
AOI Machine
Testronics 505 Machine Vision -Works off of a
known good board -Check for part
orientation -Check for missing parts
65
Final Inspection

• Bad soldering
• Contamination
• Misaligned Parts
• Wrong parts

66
Functional Test

• Tests all interfaces on the board
• Program NAND with XLoader and UBoot
• Tested pre and post burn-in

67
Burn-In

• 72 Hour Burn In
• Room temperature
• Running UBoot

68
Final Results

• 96 yield
• Kept single board penalization
• Kept assembly POP process
• Kept Profile
• Now have three suppliers of PCB
• DDI (Dynamic Details, Inc)
• MEI
• Streamline

69
REV A...Each Revision

• Only 50 units built
• Rev A...Initial Version
• Rev A1Normalized LED brightness levels
• Rev A2...Changed resistor loading options for
  S-Video
• Rev A3Lowered pull-up values for the I2C busses
• Rev A4...Lowered value of USB cap due to turn on
  issues
• Too much current drain
• Rev A5...Incorrect inductor value on TPS65950
  switchers
• Typo in BOM
• Issues to be fixed
• DC Voltage jack
• 1K pull-up on wait line
• Plated through hole issues
• Remove 4 test points
• User0 and User1 LEDS shorted

70
REV B ...Each Revision

• Rev B1Initial Release
• Fixed outstanding Rev A issues
• Rev B2USB Host not working reliably
• Removed from BOM and assembly
• Questions around the layout
• Rev B3Added a few caps back in from B2
• Rev B4...Some USB HUBS not connecting on OTG Port
• Noise level too highAdded a capacitor onto USB
  power rail
• Rev B5...Serial Port disconnects after a while
• Removed a capacitor on the 32KHZ Clock
• Rev B6PCB spin to change package of U9 and U11
• Issues To be Fixed
• USB Host

71
Big Issue Emerges REV B

• Failures of the serial port and EDID
• Traced to a bad package on TXS0102
• Delamination issue
• 60 Failure rate
• Started 72 burn in process to weed out
• Still have 1 making it out
• Only solution is to move to a different package

72
1st Rev C Proto

• Attempted to fix USB Host
• Significant improvement
• Still not 100
• This version became the board for Rev B6
• Replaced U9 and U11 package
• Changed Revision to B1

73
2nd Rev C

• PCB In FAB
• Moved USB Host to port 2
• Aligned with other platforms
• Success on Port2
• Added native LCD access
• Lot of demand for this feature
• Ability to interface to different displays with
  small paddle boards (LVDS, TTL, etc.)
• Kept current expansion header

74
Support Process
75
Key Issues from Users

• Serial Port Issues
• Confusion over cables
• Serial port failures
• OTG to host connectors
• Need special cable
• Brought on by lack of USB Host port
• Connecting another video output
• LCD

76
The Returns
77
RMA By Subsystem

• Audio..0
• DVI-D..2
• Serial Port..19
• Memory...0
• S-Video...0
• SD/MMC.0
• USB OTG...0
• User Issue..9
• PMIC3
• Nothing Found...1

34 Total..1.7 Serial Port..1.0
78
The Serial Ports

• 60 of returns for this reason
• Issue with a small number of boards due to part
  problem
• This issue masks other real issues
• Wrong IDC Cable
• Wrong SERIAL Cable
• Wrong terminal setup
• Well, it works on that other board?

79
Implementing the Lessons LearnedWhats Next?
?
80
Changes Being Made

• Documentation
• Troubleshooting Section for Serial Port
• Pictures
• Step by Step process
• Diagnostic SW
• New Translator Package
• PCB Change REV B6
• LCD Interface (Rev C)

81
Design Material Available
http//beagleboard.org/hardware/design

• Schematics
• PDF Version http//www.beagleboard.org/upload
  s/BBSRM_B5.pdf
• OrCAD Version http//www.beagleboard.org/uploads/
  BEAGLE_ORCAD_B4.zip
• BOM
• Excel http//www.beagleboard.org/uploads/Beagle_BO
  M_B4.xls
• PCB Files
• PCB Gerber Files http//www.beagleboard.org/upload
  s/Beagle_Allegro_B.zip
• PCB Allegro Files http//www.beagleboard.org/uploa
  ds/Beagle_Gerbers_B.zip
• Viewers
• PDF
• OrCAD
• Gerber
• Allegro
• PCB Design Guideline http//focus.ti.com/dsp/docs
  /dspsupporttechdocsc.tsp?sectionId3tabId409fam
  ilyId1526abstractNamespraav1
• System Reference Manual http//www.beagleboard.org
  /uploads/BBSRM_B5.pdf
• Rev B6 on its way

82
Can you do OMAP3?

• Take the BeagleBoard and build it
• Verifies if the PCB Fabricator can do it
• Verifies if the Assembly house can build it
• All material needed is provided
• All needed SW is available
• Diagnostics
• Kernels
• Distributions

83
Thanks To

• Contract Manufacturer
• CircuitCo www.circuitco.com
• PCB Fabricators
• Marcel Electronics mei4pcbs.com
• Dynamic Details www.ddiglobal.com
• Streamline www.streamlinecircuits.com
• PCB Layout
• ION Design www.iondesign.com
• Keith Gutierrez , Texas Instruments

84
Questions and Discussion