Add Title here by liaoxiuli


									June 2005
NTD Developments Overview
     For MNRF Symposium 7 June 2005

           Presented by: Colin Jacka
                 7 June 2005
Requirements from MNRF Grant                     

 Stated aims at 2002 MNRF initiation:
    To develop multi-beaming antenna technology
    Advanced optical signal transport
    Advanced signal processing schemes
    Developing interference mitigation techniques
    Integrated into an operating instrument which would
     benefit the development path towards the SKA
    Would make use of project deliverables from other
     MNRF-funded projects eg CABB, MMIC, SKA Siting
MNRF Progress to date                                      

 Original NTD Project Plan
    Choose NTD concept by 30 June 2004
    From that point on, the effort has been on FPAs using
     parabolic dish antennas
      • Revised Preliminary NTD Project Plan 30 September 2004
      • Roadmap towards SKA maximising Australia’s participation
          • NTD providing prototyping and design for xNTD, because
          • xNTD could be a typical SKA station
SKA Roadmap: The route forward              

                   extended New Technology Demonstrator
                    Outline
                       20 x 15m dishes at WA site
                       Complete by 2008
                       MNRF/WA/CSIRO support
                    Goals
                       Maintain Australian radioastronomy at
                        forefront of world science
                       Maximise influence in SKA project
                          • Science
                          • Technology
                          • Siting
                       Deliver outcomes for industry
                          • FPA/Digital beamforming
                          • Data-mining
NTD and xNTD - Proposals                                

 extended New Technology Demonstrator

 Objectives:
   Key item in the strategy of maximising Australia’s
    participation in the SKA (objective in CSIRO’s 2003-
    2006 Strategic Plan, and ATNF’s SKA Roadmap)
   To build a new world-class radiotelescope at the
    Australian SKA site
     • Telescope in its own right, with a lifetime of 5+ years,
       operated by ATNF as a National Facility
     • But also a prototype for the SKA
     • Risk mitigation for the International SKA Pathfinder
       (ISKAP), and leveraging of international funds
What difference does the x make?                           

   NTD
     Funded by existing, secured ATNF + MNRF funds
     2 interconnected dishes, 15m diameter, each with focal plane
      array, at proposed SKA site

   xNTD
     Additional funding from CSIRO, ATNF, WA State Gov, (and still
      soliciting others)
     20 dishes, 15 m diameter, arranged in one group, genuine
      micro-SKA, at proposed SKA site
     Project Plan: Design & Development Program until early 2006 is
      common for NTD and xNTD
     xNTD implementation phase from Jan 2006, as a result of
      sufficient risk mitigation in areas of antenna, FPA, digital
      beamforming and correlator design
NTD/xNTD Project Strategy
Challenges for xNTD                                              

   List of (technology) challenges that we keep in mind, and note that
    NTD should allow us to tackle some of them:
        Can we make small steerable dishes cheap enough?
        Cheap, high performance (wide band and polarization pure) FPAs?
        Cheap, high performance integrated RXs?
        No self-generated RFI from RXs (or rejection schemes)?
        How to transport signals from FPA?
        DBF (efficient, cost-effective using FPGAs)?
        Calibration with synthesized varying beam patterns?
        Correlator (a very large effort)
        Data storage & transportation
        Remote operation as a NF from East Coast of Oz?
   But also we note that:
      Many other large projects currently on-going
      Shortage of key people
So, where are we now?                     

 (time marches on, but) we have made substantial
  progress in a number of areas
 Antennas for NTD
 RXs for NTD
 FPA for NTD
 Digital H/W for NTD beamformer and design of
  xNTD correlator
 Interferometer experiment @ Marsfield
NTD Antenna System

   Presently looking at 3 alternatives to meet the challenge of
    1.   The Indian PPD dish design
    2.   New design using manufacturing techniques available in Australia
    3.   Refurbishing 2 antennas from Fleurs (for NTD)
   All are on-going investigations for xNTD purposes, but for the NTD we are
    going with the ex-Fleurs antennas
(1) Indian PPD Reflector Prototype   

              Photos from Ken Skinner of SES
(2) Reflector antenna options

  Custom-built mesh reflector using NC machine tools
      “High-tech” solution with high accuracy, good repeatability, and no
       tooling-up costs
      Local manufacture of prefabricated “flat-pack” reflector; assemble on
  Progress:
      Engineering consultants have been contacted to provide structural
       engineering analysis and comments on proposal
      Manufacturers asked to comment on manufacturability issues of our
      Bristow Laser Systems have demonstrated new CNC machinery which
       appear to be ideal for manufacture
      Advice being sought for patenting suitability of manufacturing methods
(1) Reflector antenna options

  Refurbished dishes from the former Fleurs
      Two 13.8m dishes have now been removed from Fleurs and are
       being refurbished at SES
      SES is confident that the condition and design are sound
      New antenna drive system has been designed
      2-element interferometer @ Marsfield
      Sites chosen for antennas at Marsfield, and infrastructure design
       has commenced
      Expect antennas to be on site at end of July, and we are
       preparing infrastructure to meet that date for installation
      Project milestone for the installation to be ready for experiments
       by end of October
Fleurs dishes
Focal Plane Arrays                                  

 John O’ Sullivan
    THEA tile for initial NTD experiment
    “Bunny ears” concept for increased bandwidth
 John Kot
    Access to Uni of Mass s/w
    Fundamentals, modelling & use of other s/w
 Stuart Hay
    Some other interesting ideas, eg foveated arrays
 Douglas Hayman
    Various other performance & measurement investigations
 Interest for collaboration in R&D from Canadians, Sth
  Africans, Astron, UKSKADS
FPA options

 Collaborative development of “Vivaldi” array with ASTRON /
     Appears to be quickest option for short-term demonstrator
     Have price from ASTROM for delivery of THEA tile
     Agreement with U.Mass & Astron for use of s/w for both NTD &
      xNTD development
 Attendance at FPA workshop at Astron in June
     Increase our involvement with other international FPA
 Alternate wideband arrays
     Looking at promising “rabbit ears” design
     Inherently wideband structures
     Foveated array with “natural” scaling of FoV
 Looking at wider system integration aspects of optimisation
  of FPA elements, LNAs, RX

 (200 RXs per dish equipped with FPA)
 Prototype receiver design has been completed,
  development and testing
 Some modifications discussed for use at “congested radio
  spectrum” at Marsfield
 Separate LNA for Tsys requirements: being undertaken
Digital Signal Processing                                     

   Designs have been refined such that they can efficiently handle:
    1.   A 20-antenna xNTD using 10x10 element dual polarisation FPAs
    2.   A 500-antenna LFD (MIT Haystack) @ Mileura
    3.   A SKAMP-3 with 96 antennas
    4.   An LFD with only 48 antennas (LFD fallback @ Mileura)
   All require a correlator
   NTD, xNTD and SKAMP require a digital beamformer for each
   While LFD requires a digital receiver
   Whitepaper shows commonality in the NTD beamformer and the
    LFD RX
   Sth Africans also interested, but assessing other options as well
Digital Signal Processing                                                     

   Digital Beamformer design: one for each polarization on each
         • A considerable amount of h/w required to meet xNTD specs, but NTD is
           being used to mitigate risks,
         • should only build h/w where it has value (rather than simulation, theory,
           paper designs), eg need inputs from all RXs for FPA analysis, but we don’t
           really need to have 48 beams on NTD. So NTD beamformer to have less
   Prototype NTD beamformer has 24 MHz bandwidth, and 24 inputs
    (final DBF requires 250 MHz bandwidth and 200 inputs)
   Status:
      Daughter Boards are being populated
      Motherboard (supporting 6 daughter boards) is being manufactured
      1st stage PFB (for full DBF) has been designed & debugged in Xilinx
NTD beamformer                                                                          

    ~512 MS/s Real or      16 channel
                                                      Beamformer   48 16MHz
    ~256 MS/s Complex      Polyphase
                                                       ~48 beam    1k Channel      48 beams
                        Output 16x16MHz
                                                        16 MHz     Polyphase       256 MS/s
       Inputs from                        Routing                               (4R,4I) complex
       100 Vivaldi                        network                                 100 Gbits/s
          feeds                           Outputs
                                          16MHz                                   to central
                                          all feeds                             correlator and
    ~512 MS/s Real or      16 channel
                                                      Beamformer   48 16MHz      beamformer
    ~256 MS/s Complex      Polyphase
                                                       ~48 beam    1k Channel
                        Output 16x16MHz
                                                        16 MHz     Polyphase

 Input data rate for one polarisation from 10x10 FPA from one
  xNTD antenna is 100x256Mx2x8 = 409.6 Gbps
 Output data rate is reduced by 4 = 100 Gbps
 (but 20 antennas and 2 polarisations means an output data
  rate of 4 Tbps to the correlator)
Possible xNTD correlator                                                                     

                                        Dual Rocket               Correlator Board
                                         I/O Links
                                         2x8 Gb/s
                                                                                         DDR2   DDR2
                           8 antenna                    Beam 1 HF
                                                                         HF Correlator
                            2 beam                    Router and data
                                                                            Beam 1          LTA
                           Router and                    serialiser
                                                                        2 x XC4VSX35
                             Buffer                     XC4VFX20                                DDR2

                                                                                         DDR2   DDR2
                                                        Beam 1 LF
                                                                         LF Correlator
                                                      Router and data
                                                                            Beam 1          LTA
                           8 antenna                                    2 x XC4VSX35
                                                        XC4VFX20                                DDR2
                            2 beam
                           Router and
                                                                                         DDR2   DDR2
                             Buffer                     Beam 2 HF
                                                                         HF Correlator
                                                      Router and data
                                                                            Beam 2          LTA
                                                                        2 x XC4VSX35
                                                        XC4VFX20                                DDR2

                                                                                         DDR2   DDR2
                           8 antenna                    Beam 2 LF
                                                                         LF Correlator
                            2 beam                    Router and data
                                                                            Beam 2          LTA
                           Router and                    serialiser
                                                                        2 x XC4VSX35
                             Buffer                     XC4VFX20                                DDR2

    Need “smart” design to avoid the routing of the data strangling the design
    Design approach is to use a correlator cell that minimises i/o requirements
    1 Virtex 4 XILINX FPA chip processes 48 frequency channels for all 20
     xNTD antennas simultaneously.
    Full xNTD correlator (48 beams) requires:
         24 correlator boards, each with 16 mid-sized XILINX chips
         72 Buffer boards, each with 5 Xilinx chips and 16 memory chips
Some Influential Milestones                    

 Endorsement of SKA Roadmap by ASKACC (done)
 Australian SKA site selection [NSW/WA] Nov 04 (done)
 Approval by AABoM and DEST for revised NTD plan (done)
 Australian inter-departmental SKA Steering Committee
  report Feb-Mar 2005 (done)
 Secure additional CSIRO funding (done)
 GO/NO-GO Decision for xNTD at Dec 2005 (Mar 2006)
 SKA international site selection end 2006
Focal Plane Arrays for the
    New Technology

  John Kot, Stuart Hay, Nasiha Nikolic, Doug Hayman,
                   Christophe Granet
                      NTD FPA system          

 Individual FPA element has
 high TA due to spill-over
                                    Analogue receiver

Mutual coupling and
reflections in the array +    Quantization & digital
reflector system              beamforming to generate
                              low TA antenna beams
Focal field overlaid with array                                         

         800MHz 1.9° scan                                         1800MHz 2.6° scan
   Contours at -6 and -10dB
   Number of elements required to form a beam estimate:
        About 8 elements in both cases should give a basic beam
        More are required to clean up the beam and increase efficiency
   For a full critical sampling at 1800MHz, about 16 elements are needed at
Basic FPA Operation                                   

 The diameter of the focal spot          f
 Reflector beamwidth   / D
    To 1st order, FoV is constant across the band
    For a wideband FPA, to form high efficiency beams at the
     low-frequency end of the band requires summing inputs
     from many elements
 For off-axis beams, focal spot offset increases with
  f/D, while coma decreases
    For a given FPA diameter and FoV there is an optimum f/D
 Element spill-over temperature is high: receiver gain
  must be reduced to avoid clipping in A/D converter
FPA Size vs. F/D                       

 Optimum F/D depends on quality of edge beams
 For half scan of 4°:
     3dB loss 0.35 – 0.4
     1dB loss 0.4 – 0.5
Low noise operation with a wideband,
uncooled FPA                                 

 FPA element impedance is strongly determined by
  array effects (tendency for large variations with
 For a small FPA of identical elements, the element
  impedances are all different (modulo symmetry)
 LNA noise contributions: self noise + coupled
  radiated noise from all other LNAs
 We need to do a lot of work to understand how to
  achieve optimum low noise operation – it is far
  from straight forward
Ongoing efforts towards an NTF FPA                     

 CSIRO has signed an agreement with U. Mass.
  and ASTRON for joint development of a ”Vivaldi”
  FPA for 800MHz – 1.8GHz
    Avoid “re-inventing the wheel”
    Mitigate a major risk with the NTD project
 Development of a model for a narrow band “egg-
  crate” array of dipole feeds with a reflector
    Gives us a simple, analytic model to study representative
     FPA effects such as: noise coupling; element spacing;
     offset between dual-polarized array elements; interaction
     between array and reflector (“cavity effect”)
    Allows us to explore more of the parameter space than
     currently possible with a finite element / boundary
     element computational model
Ongoing efforts towards an NTF FPA (2)                 

 Investigation of alternative array elements & tilings
    Non-ideal aspects of uniform Vivaldi arrays: polarization
     and large excursions in element impedance
    Other elements may be better, e.g. arrays derived from
     self-complementary screens have very large impedance
    Non-uniform arrays may compensate for “small array”
     effect, or offer much larger bandwidths (foveated arrays)
 Close cooperation between antenna, RF, and DSP
  engineers to understand system aspects of FPA
    Investigation of options such as balanced vs.
     conventional LNA
    Understanding of how array & reflector choices affect
     beamformer complexity
Ongoing efforts towards an NTF FPA (3)                 

 Development of a comprehensive LNA + FPA +
  reflector model, using a generalized scattering
  matrix model
    Will allow us to investigate beamforming for optimization
     of A / Tsys taking into account effects such as noise
     coupling and cavity effect, using accurate models for FPA
     and reflector and / or measured results
The End

To top