CERN–2008–003 14March2008 ORGANISATION EUROP(cid:131)ENNE POUR LA RECHERCHE NUCL(cid:131)AIRE CERN EUROPEAN ORGANIZATION FOR NUCLEAR RESEARCH CERN Accelerator School Digital Signal Processing Sigtuna, Sweden 31 May–9 June 2007 Proceedings Editor: D. Brandt GENEVA 2008 CERN–320copiesprinted–March2008 Abstract These proceedings present the lectures given at the twenty-first specialized course organized by the CERN Accelerator School (CAS), the topic being Digital Signal Processing. The course was held in Sigtuna,Sweden,from31May–9June2007. Thisisthefirsttimethistopichasbeenselectedforaspe- cialized course. Taking into account the number of related applications currently in use in accelerators around the world, it was recognized that such a topic should definitively be incorporated into the CAS seriesofspecializedcourses. Thespecificaimofthecoursewastointroducetheparticipantstotheuse and programming of Digital Signal Processors (DSPs) and Field Programmable Gate Arrays (FPGAs) evaluation boards. The course consisted of lectures in the mornings covering fundamental background knowledge in mathematics, controls theory, design tools, programming hardware platforms, and imple- mentationdetails. Intheafternoonsthestudentssplitintotwogroupswithpeopleworkinginpairs. One group worked on problem solving using a PC-based workstation connected to a DSP evaluation board while the second group performed the same exercise on PCs connected to a FPGA evaluation board. Half-way through the school, those working on DSPs moved to FPGAs and vice versa. The problems to be solved were of increasing difficulty and all of them were directly related to real and practical realizationscurrentlyinuseintheacceleratorfield. iii iv Foreword The aim of the CERN Accelerator School (CAS) to collect, preserve and disseminate the knowl- edgeaccumulatedintheworld’sacceleratorlaboratoriesappliesnotonlytogeneralacceleratorphysics, but also to related sub-systems, equipment, and technologies. This wider aim is achieved by means of specializedcourses. For2007,thetopicofthecoursewasDigitalSignalProcessing(DSP)andwasheld attheSigtunahÃu˝jdenHotel,Sigtuna,Sweden,from31May–9June2007. Thecoursewasorganizedin collaborationwiththeUppsalaUniversity,Sweden. ThiswasthefirsttimeDigitalSignalProcessingwasusedasatopicforaCASspecializedcourse. However,whenconsideringthenumberofrelatedapplicationscurrentlyinuseinthefieldofaccelerators, itwasfeltthatitwastherighttimetoaddressthisspecifictopic. Theorganizationofsuchacoursewas arealchallengeinitself,sinceitimpliedfindingandtransportingcomputerstoSweden,negotiatingthe availability of software licences for the duration of the course, negotiating the acquisition of DSP and FPGAevaluationboards,andpreparing50computerstobecompatiblewiththeproposedexercises. Itis a pleasure to express my thanks to CERN, the companies Synplicity, Mathworks and Xilinx, as well as theTexasInstrumentsWorldwideUniversityProgram,andtheXilinxUniversityProgramfortheirkind support,withoutwhichthecoursewouldnothavebeenmadepossible. To ensure that the required material was made available was certainly a mandatory pre-requisite, but the elaboration of the course and, in particular, the hands-on courses in the afternoon sessions were nolesschallenging. IwouldparticularlyliketoacknowledgetheoutstandingcontributionofourCERN colleagues Hermann Schmickler, Maria-Elena Angoletta and Javier Serrano for their contacts with the relevant industry and the high-quality afternoon courses they prepared and supervised, John Evans for his hard work in obtaining the software licences and Edwige Bournonville for the perfect preparation of the computers as well as their maintenance during the course. Similarly, I would like to thank Leif Gustafsson and Pawel Marciniewski (Uppsala University) for their highly appreciated contribution as tutorsduringtheafternooncourses. It is also important to thank the Local Organizing Committee (Professor Tord Ekelof, Dr. Volker Ziemann and Ms. Inger Ericson) for obtaining financial support from the Swedish Research Council (Vetenskapsraadet)tohelpCASwiththeorganizationofthecourse. As always, the backing of the CERN management, the guidance of the CAS Advisory and Pro- grammeCommittees,theattentiontodetailoftheLocalOrganizingCommitteeandthemanagementand staffoftheSigtunahÃu˝jdenHotelensuredthattheschoolwasheldunderoptimumconditions. Very special thanks must go to the lecturers for the enormous task of preparing, presenting, and writinguptheirtopics. Finally,theenthusiasmoftheparticipantswhocamefrommorethan23differentcountriesaround theworldwasconvincingproofoftheusefulnessandsuccessofthecourse. Itismypleasureandprivilegetothankmostsincerelyallthosepersonswhohelpedmetomakethe courseasuccess,includingBarbaraStrasser,SuzannevonWartburg,andtheteamoftheCERNScientific TextProcessingServicefortheirdedicationandcommitmenttotheproductionofthisdocument. DanielBrandt CERNAcceleratorSchool v Saturday 9 June Buses to airport February 2007 Friday 8 June Real Time Control of Beam Parameters II M. Dehler Controls Integration II T. Shea A K BI Application II U. Raich Outlook or Seminar H. Schmickler Parallel Lab course (DSP/FPGA) Group B/A Discussion Session D. Brandt A K Parallel Lab course (DSP/FPGA) Group B/A DINNER Thursday 7 June Real Time Control of Beam Parameters I M. Dehler RF Application II T. Schilcher F F E E B R E Controls Integration I T. Shea BI Application I U. Raich L U N C H Parallel Lab course (DSP/FPGA) Group B/A Parallel Lab course (DSP/FPGA) Group B/A F F E E B R E Parallel Lab course (DSP/FPGA) Group B/A Special Dinner P) weden Wednesday 6 June Math IV M. Hoffmann RF Application I T. Schilcher C O Transverse Feedbacks M. Lonza Longitudinal Feedbacks M. Lonza Parallel Lab course (DSP/FPGA) Group B/A Parallel Lab course (DSP/FPGA) Group B/A C O Parallel Lab course (DSP/FPGA) Group B/A DINNER SS ME sing (Dgtuna, Tuesday 5 June E X C U R S I O N PROGRAMDigital Signal ProcesMay – 9 June 2007, Si Monday 4 June Math III M. Hoffmann Control Theory III S. Simrock Introduction to DSP III M. Angoletta Introduction to FPGA III J. Serrano From Analog to Digital III J. Belleman Parallel Lab course (DSP/FPGA) Group A/B Parallel Lab course (DSP/FPGA) Group B/A DINNER 1 3 e e Sunday 3 June Math II M. Hoffmann Control Theory II S. Simrock R E A K Introduction to DSP II M. Angoletta Introduction to FPGA II J. Serrano H From Analog to Digital II J. Belleman Parallel Lab cours(DSP/FPGA) Group A/B R E A K Parallel Lab cours(DSP/FPGA) Group A/B DINNER Saturday 2 June Math I M. Hoffmann Control Theory I S. Simrock C O F F E E B Introduction to DSP I M. Angoletta Introduction to FPGA I J. Serrano L U N C From Analog to Digital I J. Belleman Parallel Lab course (DSP/FPGA) Group A/B C O F F E E B Parallel Lab course (DSP/FPGA) Group A/B DINNER Friday 1 June Types of Accelerators and Specific Needs I T. Shea Types of Accelerators and Specific Needs II T. Shea High Level Modeling Tools I J. Evans High Level Modeling Tools II J. Evans Introduction to Afternoon Courses H. Schmickler Parallel Lab course (DSP/FPGA) Group A/B Parallel Lab course (DSP/FPGA) Group A/B Welcome Drink me 30 30 30 30 00 00 00 00 30 30 30 30 00 00 00 Ti 08: 09:09: 10: 11: 12:12: 13: 14: 15:15: 16: 17: 18:19: vi Contents Foreword D.Brandt ................................................................................ v High-levelmodellinglanguages J.Evans .................................................................................. 1 Digitalsignalprocessingmathematics M.Hoffmann ............................................................................ 11 Controltheory S.Simrock ............................................................................... 73 Fromanalogtodigital J.Belleman ............................................................................ 131 Digitalsignalprocessorfundamentalsandsystemdesign M.E.Angoletta ......................................................................... 167 IntroductiontoFPGAdesign J.Serrano .............................................................................. 231 RFapplicationsindigitalsignalprocessing T.Schilcher ............................................................................ 249 Multi-bunchfeedbacksystems M.Lonza ............................................................................... 285 Realtimecontrolofbeamparameters M.Dehler .............................................................................. 331 Beamdiagnostics U.Raich ............................................................................... 361 Controlsystemintegration T.J.Shea ............................................................................... 385 LHCtechnologicalchallenges: useofdigitalsignalprocessorsinthepowerconvertersfortheLHC particleaccelerator H.Schmickler .......................................................................... 411 ListofParticipants ........................................................................ 429 vii High-level modelling languages J. Evans CERN, Geneva, Switzerland Abstract This paper gives an introduction to the latest developments in modern electronic design methodology. It will give a brief history of the evolution of design software in an attempt to explain the seemingly haphazard development up to the present-day situation. 1 Introduction This paper gives an introduction to the latest developments in modern electronic design methodology. It will give a brief history of the evolution of design software in an attempt to explain the seemingly haphazard development of these tools to become what is now known as the Electronic Design Automation industry. The emphasis is to provide a common background to all school attendees with specific references made to the tools used during the week. The frenetic pace of advance in the electronics industry is not expected to change in the near future. According to the research firm Gartner, there is no slowdown in the expected revenue growth for the Field Programmable Gate Array, Application-Specific Integrated Circuit and Application- Specific Standard Product markets for the next few years [1]. FPGAs look set to maintain the highest compound annual growth rate (almost 20%) as more and more applications adopt this solution due to their increasing functionality and decreasing costs and the often prohibitive development costs for ASICs and ASSPs. (ASSPs are essentially ASICs dedicated to a specific application market and sold to more than one user.) The total 2007 revenue from three families is expected to reach almost 100 billion US dollars. Modern-day electronic design tools have evolved for a very good reason. The increasing size, speed, and complexity of these new products and the increasing pressure on bringing them to market quickly means that design targets would be impossible to reach without help from software programs. Tool development has been continuous over the last thirty years. What traditionally was based on the two main axes of analog or digital designs has now been expanded to include software additions and we shall see how these previously distinct areas are becoming increasingly blurred and merging. 1.1 A brief history Before electronic design automation, PCBs were designed by hand and manually laid out. Perhaps more surprisingly, this was also true for IC design. The first programs developed were drafting software that included some digitizing capabilities with different companies focusing on specific areas of the market. Calma was a major player in the area of IC manufacture (its GDSII format still exists) while other essentially similar products targeted the PCB sector. While invaluable in laying out a board or IC, there was still no link between the design process and the eventual layout. Various large companies developed their own in-house software and methodology solutions and some of these teams were eventually spun out as separate companies in the beginning of the 1980s. Various pioneers of the EDA industry date from this period: Mentor Graphics (original parent company Tektronix), Daisy Systems (Intel) and the forerunner of Cadence Design Systems, Valid Logic Systems (spawned from Hewlett Packard and the Lawrence Livermore Laboratory). 1 J. EVANS The need to simulate the transistors being laid out for all ICs was one of the inspirations for the development of the Simulation Program with Integrated Circuit Emphasis (SPICE) software at the University of California, Berkeley. The main driving forces behind the original SPICE program are credited to Larry Nagel working under the supervision of Donald Pederson. The work was based on the CANCER program and was one of the many such projects funded by the United States Department of Defense. Pederson insisted that the newly written SPICE be far enough removed from the original CANCER such that any distribution restrictions could be removed and that the program be put in the public domain. The first public presentation of SPICE1 at a conference was in 1973 [2]. This was a relatively limited-feature program based on nodal analysis (leading to problems in describing some circuit elements, particularly inductors) and fixed time-step transient analysis (this aspect will be discussed in more detail later in this paper). SPICE2 was introduced in 1975 and offered improvements such as more inherent circuit elements, variable time-step analysis, and solutions based on modified nodal analysis. These were significant advances and ultimately led to the program’s widespread adoption. SPICE2 was further developed until 1983 with the introduction of SPICE 2g6 which was the last FORTRAN version of the program. SPICE3 was written in C and was introduced in 1989. The insistence on SPICE’s open-source development (and famously, program availability for the cost of a magnetic tape) led to its becoming widely distributed and used as the basis for countless other simulators in the academic, industrial, and commercial worlds. Well-known commercial versions still existing today include HSPICE (now owned by Synopsys) and PSpice (currently owned by Cadence Design Systems). Most ‘digital’ design tools also being developed during the same period were largely based on schematic-driven design entry. One of the next major advances was the development of text-based Hardware Description Languages (HDLs) describing design behaviour at the Register Transfer Level. RTL describes the operation of synchronous digital circuits where signals are treated before being clocked into hardware registers. HDLs allowing circuits to be described using RTL provide a much higher level of abstraction than at transistor or at gate level. Since the strengths and advantages of this level of abstraction were realized, tools were subsequently developed to allow direct and automatic gate synthesis from the HDL circuit description. This was a major step forward in the EDA industry and allowed a huge advance in design complexity and reduced time to market. Two major HDLs became prominent during the mid-80s. Verilog is a proprietary language defined by Gateway Design Automation. This was an industry changer and is still the prominent language in certain design areas. Written by Phil Moorby in around a year, it quickly became the de facto standard for ASIC design along with the Verilog-XL simulator. Its impact was reinforced when Synopsys introduced software permitting direct gate-level synthesis from the Verilog netlist. This allowed a direct line from design description to simulation and layout in one integrated flow. Gateway was eventually acquired by Cadence Design Systems during 1989. With the increasing market penetration of VHDL (described below), Cadence decided to make the language generally available through the Open Verilog International organization. The language was later submitted to the IEEE for standardization and this was eventually accorded as IEEE Standard 1364-1995. Subsequent developments of Verilog have culminated with the release of SystemVerilog. The Very-High-Speed Integrated Circuits Hardware Description Language (commonly abbreviated to VHDL) was developed for the United States Department of Defense. It was becoming impossible to reproduce reliably the functionality of technically obsolete equipment as its specifications were insufficiently documented. VHDL was conceived as an attempt to overcome this limitation by defining a language that could formally and completely describe an object’s specification. It was based on the existing ADA language syntax so as to benefit from already well- proven concepts. As with Verilog, the idea of being immediately able to simulate and synthesize the 2
Description: