October 2005ALMA Cost Review1 A Glimpse of Technology Advances for the ALMA Project. ALMA Correlator Hunt for Molecules Paris, September 19-20 2005 Alain.
Post on 29-Jan-2016
215 Views
Preview:
Transcript
October 2005 ALMA Cost Review 1
A Glimpse of Technology Advances for the ALMA Project.
ALMA Correlator
Hunt for Molecules
Paris, September 19-20 2005
Alain Baudry
Université de Bordeaux 1, Observatoire Aquitain des Sciences de l’Univers
October 2005 ALMA Cost Review 2
ALMA, a World Array
*NAOJTokyo
*ESO & Institutes*NRAO
CV & Socorro
*Santiago*ALMA site
October 2005 ALMA Cost Review 3
ALMA Site
Santiago
Paranal
La Serena
October 2005 ALMA Cost Review 4
ALMA Site (5000-m, Array Operations Site -AOS-)& Site Transmission
October 2005 ALMA Cost Review 5
ALMA Array
October 2005 ALMA Cost Review 6
Staff work at the Operations Support Facility (OSF) at an elevation of ~ 2800 m.
Astronomers will not normally visit the OSF.
A new road connects the high site with the San Pedro/Tocanao highway.
The OSF is < ~ 1 hour drive from San Pedro.
Operations Support Facility (OSF)
October 2005 ALMA Cost Review 7
ALMA Overview
• ALMA is a European & North American bilateral project + Japan to construct and operate an array of
- 64 12-m antennas + ACA (12x7-m + 4x12-m)
• Reconfigurable array (~ 150 m to 14 km) to image astronomical objects
• Wavelength coverage, ~ 3 mm to ~ 300 µm with cooled low noise SIS Rx
• Final IF = 4 x 2-4 GHz and 2 polarizations per antenna=> 8 digitizers per antenna working at 4 Gsample/s & digitized signal transmitted through optical fibers
• Bandwidth selection with Digital Filters: 2 GHz to 62.5 & 31.25MHz
• Highly flexible correlator: 8192 spectral points (for each pair of BB or in 1BB )
October 2005 ALMA Cost Review 8
Many Technical Challenges
• Ambitious Science requirements: extreme sensitivity (a few µJy in 1h in cont., or map spiral galaxies at z=2 in CO and C lines), few kHz resolution, accurate calibration of phase and intensities, etc.
Technical requirements difficult to meet
Examples- Antennas: surface accuracy ~ 20-25 µm, max pointing error ~ 0.6”/ref source at 2°, fast switching
- Rx noise ~ a few quantum limits- Highly stable LO: photonic LO with highly stable lasers beating in
photomixers - Digitization early inthe processing chain: at antennas- High speed samplers: working at 4 Gsample/s input up to 4 GHz- Digital filters: new practice in RA- New correl. chip: 4 klags, 1 lag = 2-bit Xer at 125 MHz, ~ 2 million gates
October 2005 ALMA Cost Review 9
Many Technical Challenges
• Many stability or synchronization problems to be solvedPhase coherence throughout the array achieved with a stable fiber to carry the LO reference signals to the antennas (e.g. 3 degree phase at 300 GHz => 4 10**-10 stability for transport over 14 km)
Fibre length stability achieved by both passive (thermal insulation) and active control (round-trip length control).
Frequent atmospheric phase calibrations (overall phase stability)
• Production of many (thousands) sub-systems, not traditional in RA => industrial involvement at both developt & production phases
• Considerations on cost and reliability (remote site) present troughout the project
Design of Subsystems => compromises between fully innovative components (requiring industrial interest for mass production) and « components off the shelf »
October 2005 ALMA Cost Review 10
System Overview
• Focus on European BackEnd deliverables– Photonic LO
– Digitizers
– TFB cards in Correlator
• Overview of ALMA correlator
Front-EndPhotonic
Local Oscillator
ANTENNA
Data Encoder12*10Gb/s
12 Optical Transmitters12->1 DWD Optical Mux
Digitizer8* 4Gs/s -3bit ADC
8* 250 MHz, 48bit out
IF-Processing(8 * 2-4GHz sub-bands)
Fiber Patch-PanelFrom 270 stations to 64 DTS Inputs
Optical De-Mux
Digital De-Formatter
Correlator
Technical Building
Tunable Filter Block
Central PhotonicLocal Oscillator
30-959 GHz
October 2005 ALMA Cost Review 11
Photonic LO OscillatorRAL development
Fiber LaserFiber Laser
Optical CombGenerator
Optical CombGenerator
PhaseLock Loop
PhaseLock Loop
ExternalCavity Laser
ExternalCavity Laser
Master Laser
Slave Laser
Round-TripServo
Round-TripServo
Round-Trip PhaseCorrection
CENTRAL CONTROL ROOM
125MHz Ref
PhaseLock Loop
Gunn DiodeOscillator
AOM
0.1-25kmBuried Fibre
High Freq.Optical Ref.
IF Out 4-12GHz
72-84GHz
CryogenicReceiver
SIS Mixer
HEMTAmplifier
ANTENNA RECEIVER CABIN
3
Photomixer Block
2 lasers beating in photomixers => signal up to 140GHz and multiplied up to produce the 1st LO frequencies
October 2005 ALMA Cost Review 12
Digitizer & Formatter + Mux to Fiber
• Numérisation à 4GS.s-1 sur 3 bits / 8 niveaux
• 96Gb.s-1 d’information
• Transmission par fibre optique à 12 x 10Gb.s-1/antenne
October 2005 ALMA Cost Review 13
Digitizer Consortium, Assembly & Test Equipment Digitizer custom chips
L3AB, Bordeaux
Concepteur des ICs :
IXL & Obs Bordeaux
Fondeur : STmicroelectronics
Assembleur : Solectron-E
PCB : Atlantec
Banc de test et Clks : IRAM
DG Sub-Assembly
10MHzSynchro
DGS#1
DGD#1
DGS#2
DGD#2
BB#1Input
BB#2Input
4GHzClockInput
250MHzClockInput
FPGA
4 GHz
250 MHzPC and Plug-In Digital
Pattern Acquisition Board
Anti-Aliasing filter
Noise Generator
RF Generators
DTEC Board
250MHzClockInput
Can be replaced by DGCK
October 2005 ALMA Cost Review 14
ALMA sampler, 3 bits, 8 levels
• CAN : structure de type flash
• BiCMOS 0.25µm en SiGe (2.5V)
• Entrée : bruit gaussien 2-4GHz
• Horloge : 4GHz
• Sorties LVDS : 3 (x2) => 4Gsamples/s
Packaging• VFQFPN 44 7x7x1• Ball-bonding
Eléments parasites - Bonding : inductance ~1nH/mm- Pad boîtier : ~0.3pF- Pad puce : ~0.08pF
4 GHz
0 dBm
Clockbuffer
VH
VL
+-
+-
+-
+-
+-
+-
+-
Amplifier
OTA
Bandgap
2-4 GHz
D2
D1
D0Output buffer
Output buffer
Output buffer
D D
DD
DD
DD
DD
DD
DD
Adapter amplifier
D
D
D
D
D
D
Encoder
October 2005 ALMA Cost Review 15
ALMA Sampler: Subassembly, PCB technology, Eye diagram
PCB• Circuit multicouches (6 layers)
• Substrat : RO4350 / pre-preg RO4450, collé sur un drain thermique pour écouler les charges thermiques.
CMS
Drain thermique
Vddx
Gnd
Gnd
Gnd
HSStripline
50Ohms 100Ohms diff.
4450
4350
4350
4350
4450Diagramme de l’œil : bit de signe
October 2005 ALMA Cost Review 16
ALMA demultiplexer
PCB• Circuit multicouches (10 layers)
• Substrat : RO4350 / pre-preg RO4450, collé sur un drain thermique pour écouler les charges thermiques.
Diagramme de l’œil : bit d’une sortie
CMS
Drain thermique
Stripline
100Ohms diff.
4450
4350
4350
4350
4450
100Ohms diff.
100Ohms diff.
4450
4450
4350
4350
50Ohms
50Ohms
Gnd
Gnd
HS
Gnd
Vddx
Gnd
Vddx
Gnd
HS
HS
October 2005 ALMA Cost Review 17
Digitizer Test Equipment
DG Sub-Assembly
10MHzSynchro
DGS#1
DGD#1
DGS#2
DGD#2
BB#1Input
BB#2Input
4GHzClockInput
250MHzClockInput
FPGA
4 GHz
250 MHzPC and Plug-In Digital
Pattern Acquisition Board
Anti-Aliasing filter
Noise Generator
RF Generators
DTEC Board
250MHzClockInput
Can be replaced by DGCK
DG CorrélateurAUTOCORRELATOR
REQ
ACK
MASTER RESET
CLOCK
DATA[23-0]
PDF BUILDER
CONTROL
LOGIC
CLOCKENABLERST_COR
CLOCKENABLERST_PDF
X00[2-0]X01[2-0]X02[2-0]X03[2-0]X04[2-0]X05[2-0]X06[2-0]X07[2-0]X08[2-0]X09[2-0]X10[2-0]X11[2-0]X12[2-0]X13[2-0]X14[2-0]X15[2-0]
MUX
PCLK
October 2005 ALMA Cost Review 18
Correlator Overview
• Development began with MMA architecture concept at NRAO and with Second Generation Correlator (2GC) concept in Europe
• ALMA Correlator evolved to include the Tunable Filter Bank (TFB) from 2GC study => to enhance the performance by 32 in spectral resolution– Frequency demultiplexing and filtering is fully digital (TFB)
• Tasks are performed in NA and Europe (TFB); TFB tasks managed at University of Bordeaux
• Specifications– 64 antennas, X-correl coefficients for all 2016 antenna pairings – 16 GHz BW, full Stokes polarization analysis, 3.8kHz resolution– High flexibility
• Deliverables: total of 32 racks with over 2500 PCBs – Station electronics and filetring cards– Baseline electronics cards
• Total power dissipation = 170kW
October 2005 ALMA Cost Review 19
Digital Filtering (TFB)
Digital Filters convolve digitized signals with numerical tap weights∑ Xn-i * piXi = 2- or 3-bit input samplespi = 8-bit (or more) taps loaded from outside
Convolution is implemented in FPGAs offering RAMs, DSP blocks (multipliers or adders), PLLs
• TFB demuxes each 2 GHz input band into 32 sub-bands– each subband or groups of subbands are frequency centered as required
by the astronomer
October 2005 ALMA Cost Review 20
Digital LOand Mixer
x
DDSLO
FIR filter
128 taps(8 bit encoded)
Low Pass
FIR filter
128 taps(8 bit encoded)
Low Pass
4Gs/sBaseband
32* 3 bits
8 bits
FIR filter
64 taps(9 bit encoded)
Low Pass
6 bitsReal part
6 bitsImag. part
125 Ms/s
125 Ms/s
8 bits
FIR filter
64 taps(9 bit encoded)
Low Pass
Complex to RealConversion
62.5 Ms/s
62.5 Ms/s
RequantizationOutputSignal
9 bits
9 bits
InputSignal
• DDS LO + Mixers• Two FIR Filter Stages
• Output Stage : Complex to Real Conversion and Re-quantization
Digital filter core
October 2005 ALMA Cost Review 21
Digital Filter Card on Test Fixture2 Filters /1 FPGA
FIR Chip
DISTRIBUTION
Filter Core
Filter Core
Configuration
32
input bits
3*32 input bits
3*32 input bits
32
input bits
2 output
bits SB1
2 output
bits SB1
to CPLD2
Monitor
October 2005 ALMA Cost Review 22
Progress to Date
• First quadrant hardware is built and verified
• Prototype TFB cards tested with dedicated Test Fixture and other tests ongoing in first quadrant of ALMA correlator
• Close integration with Computing team; operational software is used in system testing, operating modes are being verified
• Project is on schedule and under original budget
October 2005 ALMA Cost Review 23
Operations
• Beginning with first quadrant installation by the construction team
• Responsibility for operation and maintenance will be transferred to Chilean staff, with continued support while the remaining quadrants are built and delivered
October 2005 ALMA Cost Review 24
First Quadrant
October 2005 ALMA Cost Review 25
TFB Spectral Flexibility (1)
• Example of 13 lines arbitrarily distributed in 2 GHz– Actual ‘Line Forest’ sources
exhibit even more lines ~ 20 to 25 lines per GHz in Orion
• With TFB cards these 13 lines can be analyzed simultaneously with different spectral resolutions
October 2005 ALMA Cost Review 26
TFB Spectral Flexibility (2)
• With 8 sub-channel filters one can have (see Mode Tables)– Correlated BW = 500
MHz– Spectral resolution =
0.976 MHz with all 4 X-products and smooth result to get effective resolution of 1.95 MHz
October 2005 ALMA Cost Review 27
TFB Spectral Flexibility (3)
• Use just another sub-channel filter to obtain high frequency resolution at appropriate IF frequency around 840 MHz– Correlated BW = 62.5 MHz– Spectral resolution ~ 120
kHz with all 4 X-products
• One still has several filters and spectral channels available for other scientific purposes
October 2005 ALMA Cost Review 28
Many Correlator Mode Options
• Frequency division mode (max of 8192 spectral points in 1 Baseband per Quadrant) or Time division mode (64 spectral points in 1 Baseband per Quadrant)
• Several Modes– Select Number of Sub-channels <=> Correlated Total Bandwdth– Select Efficiency
» Sample Factor: Nyquist or Twice Nyquist (with twice less spectral points)» Improve further Efficiency: 4-bit x 4-bit Quantization (less spectral points)
– Select Polarization <=> 1 Baseband or 2 Basebands without or with (full Stokes analysis) polarization cross-products
• Resources are exchangeable among• Basebands = 1 or 2 max per Correlator Quadrant• FIR filters = 32 max per Baseband• X-products: 1, 2 or 4
October 2005 ALMA Cost Review 29
Science Cases
• Line and Dust Study in Young Stellar Objects– 1 sub-array of 16 antennas for CO line observation (230 GHz) in 62.5 MHz
bandwidth (~ 80 km/s velocity coverage) with high spectral resolution (7.6 kHz or 0.01 km/s)
– 2 other sub-arrays of 24 antennas each in 2 other frequencies to provide spectral imaging of other weaker lines in narrow total BW (~ 80 km/s) and accurate spectral index measurement of dust in broad ‘continuum’ bandwidths
• High number of channels in broad BWs essential to search for new molecular species … interferometry … glycine
• Broad BWs with several spectral channels and different receiver bands useful to measure unknown z in distant CO galaxies
• Line Survey & Imaging in Orion-like young/massive sources– Requires both broad & narrow BWs … now possible with many more channels
top related