The Fei Koala Architecture & why we hope it is future proof
Contents ! ! ! ! !
2
Introduction to Fei Why ‘future proof architecture’? Architecture model Fei Koala Architecture Conclusions
2004-1-6
Pybe Faber
5 min 5 min 10 min 10 min 5 min
Introduction to Fei History ! 1949: Philips Electron Optics sold first commercial TEM ! 1971: Fei (Field Electrons & Ions) is founded ! 1997: Fei and PEO merge Operations in ! Hillsboro (Oregon) – Main Office ! Peabody (Massachusetts) ! Eindhoven (Netherlands) ! Brno (Czech Republic) About 1600 employees worldwide, nasdaq FEIC ($26.51 on 2828-1-2004) 3
2004-1-6
Pybe Faber
Fei Products Fei Products ! Transmission Electron Microscope (TEM) ! Scanning Electron Microscope (SEM) ! Focussed Ion Beam (FIB) ! Dual Beam (DB = SEM + FIB)
4
2004-1-6
Pybe Faber
Contents ! ! ! ! !
5
Introduction to Fei Why ‘future proof architecture’? Architecture model Fei Koala Architecture Conclusions
2004-1-6
Pybe Faber
5 min 5 min 10 min 10 min 5 min
What is architecture?
!
Standardized interfaces between modules
Examples: ! Standardized dimensions (e.g. for plumbing pipes & joints or screws & bolts) ! Standardized pin-out & voltages for connectors (e.g. telephone and mains power) ! Standardized software interfaces (e.g. http, COM) 6
2004-1-6
Pybe Faber
Why architecture? ! !
!
Share infrastructure Independent development of modules (by many people) Reuse of modules
→ Reduce cost
7
2004-1-6
Pybe Faber
Why not architecture? ! ! ! !
8
Expensive to design Limitations (things that cannot be done) Education of people Modules may be more expensive / complex because they must match the architecture
2004-1-6
Pybe Faber
Why future proof? !
9
To avoid designing a new architecture » Which costs a lot of time » Which makes existing modules & tools obsolete » Which makes existing education obsolete
2004-1-6
Pybe Faber
Why new architecture? ! ! ! !
10
Desire for better performance Obsolete parts, knowledge, tools Price reduction Get rid of top-heavy old architecture (due to add-ons over time)
2004-1-6
Pybe Faber
Long-lived architectures 1795 France adopted the metric system (again in 1840) 1804 Steam locomotive (standardized tracks) 1876 Telephone 1882 Electricity (Edison’s first power station) 1969 TCP/IP
11
2004-1-6
Pybe Faber
Why long-lived?
! !
12
Prohibitively expensive to replace Good design (gradual improvements possible)
2004-1-6
Pybe Faber
Contents ! ! ! ! !
13
Introduction to Fei Why ‘future proof architecture’? Architecture model Fei Koala Architecture Conclusions
2004-1-6
Pybe Faber
5 min 5 min 10 min 10 min 5 min
UI
User Interfaces
App. Srvr.
Automation
Application Behavior
Layers
Client layer
Application layer (translate between machine and user states)
User Sessions
Automatically degauss optics
Automation layer (Optional behavior)
Hardware
Raw Behavior
Microscope Server
Synchronization layer (non-optional interconnections between models)
Source off if vacuum not OK
Detectors off if vacuum not OK
Vacuum model
e-Source model
Vacuum Firmware layer firmware (safety, performance) (PUC)
e-Source firmware (2xPUC)
Safety layer (interlocks)
Model layer (standalone, no interconnections)
Electronics layer (hardware)
Hardware Interlocks layer
Vacuum electronics (VIU, MUC)
vacuum interlock
e-Source electronics (HTSU, FGSU)
Application UI's
User Preferences and presets
Active Quad
Detector presets
Simplify machine states
Add 'commercial' states
Move stage down on vent
Spot size changes detector gain
Auto Contrast & Brightness
Pause CCD when BSD selected
Blank beam during long stage moves
Source and Optics use same HV value
Scan and acquisition use same settings
2004-1-6
Position with stage and optics
Load Lock (Load, Unload)
Stage collision prevention
Optics model
Optics electronics
Scan model
Aquisition model
Scan firmware (DSP?)
Aquisition firmware (DSP)
Scan electronics (DSPB or ISAS)
Aquisition electronics (Viper Quad or ISAS)
Detector model
Motion model
Motion firmware (Nyquist)
Detector electronics
Motion electronics (STAB)
vacuum interlock
Hardware Interconnection layer (high bandwidth and time critical connections)
14
Third parties
Scan
Video
Grouping of system behavior into logical layers. An actual implementation will have less layers than shown here, combining several layers into one. Interconnections within a single layer are forbidden: if two modules need to share data this must be done in a higher layer, and at least at the level of the Safety layer (for software) or in one of the two lowest layers (for electronics). On the left you can see a coarse behavior split between Raw (hardware specific) and Application (market specific) behavior, and how the xT Nova functionality it is split over the Microscope Server, Application Server and UI (there is still too much functionality in the Microscope Server). Note that only a few of the actual modules and interconnections are shown.
Pybe Faber
Pybe Faber 28-1-2004
Contents ! ! ! ! !
15
Introduction to Fei Why ‘future proof architecture’? Architecture model Fei Koala Architecture Conclusions
2004-1-6
Pybe Faber
5 min 5 min 10 min 10 min 5 min
Fei Architecture History !
±1985
!
±1995 1998
!
2001
!
2003
!
16
xL: first mouse-controlled SEM, P80 electronics xP: 32 bit (mainly xL electronics) Tecnai: first mouse-controlled TEM, COM, P80 electronics Quanta: new electronics, CAN bus, modularized server software, digital video Nova&Quanta3D: application server, topdown software interconnections
2004-1-6
Pybe Faber
Hardware interfaces
Electronics consoles
Microscope console
Microscope Computer
100 Mb ethernet
198.211.143.65 Ethernet Adapter
Second Ethernet Adapter
HUB
Power supplies
STage Amplifier Board (STAB)
Optics Scan (OPSS)
CAN
198.211.143.93 CCB 14
P80
TestLine (OPTD)
sta
198.211.143.82 Motion controller (Nyquist)
EOptics
GIS
HTSU
FGSU
System Services (SSIB)
HVPS
IOptics (EOCU)
Detectors (IQI)
Vacuum Interface Unit (VIU)
IGP supplies
Analog Detector outputs Analog Scan signals
17
2004-1-6
Pybe Faber
Analog Scan signals
Acquisition (Viper Quad)
Digital Scan and Patterning
Digital Scan and Patterning
Support PC
Software interfaces !
!
!
18
(D)COM interface between instrument server, application server and UI. Object Model: structured way to organize the interface, target is to have a single OM for all Fei products (will take a few years still…). Visio tool to graphically display connections between ‘bricks’ (software modules) – see next slide.
2004-1-6
Pybe Faber
Node Template : 1 = "explorer style" 2 = Square blob style
Drawing Control Panel
Bricks
filename.exe
2
Node Template Connector Template
1
Positioning Algorithm
3
Data Source
1
Connector Template : 1 = Straight line
Entity
Positioning Algorithm : 1 = Explorer small icon style 2 = Levels based on Tecnai dependancy file 3 = brick prefix
-
Source File Dim All Connections
Remove All Connections
c:\TecnaiSimpleMap.txt
Save Position Nodes
Restore Position Nodes
C:\tools\visio\Solo3n.txt
Nodes can be moved and the diagram will be updated. Right click on nodes to highlight and dim individual nodes.
Position File
Clear Diagam
Refresh All
Refresh Nodes
(duration ~1 minute!)
XT Smart Architecture Diagram OMInstrument
ObjectModel 1
OMDetectors
OMVacuum
OMDataSources 1
1
OMIBeam
1
OMSpecimenCu rrentMeter 1
1
OMPatterning
OMEBeam 1
BhvSimIBeam
OMService 1
1
1
BhvSpecimenEx change 1
1
BhvAutofunction s 1
1
BhvMaintenance 2 1
BhvSimEBeam
BhvPredict
OMPositioning
1
1
HlfUserDataServ er 1 HlfAlignments
BhvAdapter 1
1
HlfSDB 1
BhvConditions
HlfMemento 1
BhvAreaOfIntere stFinding 1
BhvMaintenance 1
1
BhvPatterning
BhvStateManage r 1
1
BhvVacuum
BhvImageForma tion 1
1
BhvBeamCorrec tions 1
BhvImageMeme nto 1
BhvNormalizatio n 1
BhvSafety
MdlPositioning
1
1
BhvDatasource 1
Mdlvacuum
MdlMagnumFIB Optics 1
1
Mdlvacruleset
1
MdlOpticsBrick
MdlLogicalDetect ors 1 MdlService
1
MdlSingleAvaBri ck 1
MdlFibgunBrick
MdlMotion
1
1
1
MdlPhysicalDete ctors 1
TadRegulators
1
MdlGunFegBrick
1
MdlImaging
1
MdlCurrentSens e 1
HalImaging
HalHwdRegulato r 1
HalRegulator
1
1
HalVacuum
halmotionAVA 1
1
HalMagnumFIBR egulator 1 HalMagnumFIBO ptics 1
HalSsib
HalDriver 1
HalRack 1
HalMux 1
HalSta 1
HalFIBGun
1
1
2004-1-6
1
HalGunHT
1
HalCcb14
19
halmotion
HalGunFEG 1
Pybe Faber
1
Contents ! ! ! ! !
20
Introduction to Fei Why ‘future proof architecture’? Architecture model Fei Koala Architecture Conclusions
2004-1-6
Pybe Faber
5 min 5 min 10 min 10 min 5 min
Hardware Trends !
!
!
21
Bare PC, only standard network connections (ethernet, possibly also USB2, firewire or CAN). Standalone modules with standardized power and network connections (ethernet or CAN). Second (‘support’) PC for 3rd party applications, post-processing, data storage (keeping the server PC load predictable).
2004-1-6
Pybe Faber
Software Trends !
!
!
!
! 22
More modular software, in a more hierarchical structure, distributed over more than one PC. As a result: focus on standardized software interfaces, with ability of ‘remoting’ (network connections). Databases for result storage (data from Fei systems, but also from 3rd party equipment). Tighter integration of 3rd party equipment, needed to automate complex tasks (focus on ‘solutions’ rather than ‘tools’). Moving to C# / .NET for application software (highest level). 2004-1-6
Pybe Faber
Lessons from Fei’s past !
!
! ! ! !
23
Complex parts are long-lived: the knowledge to design them is gone. High-risk parts are long-lived: a new design may introduce big problems. Embedded software is hardest to maintain. A new architecture MUST be linked to a product. Lean design: only standardize what is needed Don’t add bells & whistles, provide ‘convenience’ as stand-alone tools, not as part of the architecture 2004-1-6
Pybe Faber