DS4T5 Beam forming at the patch level

1
DS4T5 Beam forming at the patch level

• Primary objectives
• Compare the cost effectiveness of analogue and
  digital solutions for wide-band beam forming
• Digital Beam-former solutions
• Evaluate the cost-effectiveness of different DSP
  solutions
• Demonstrate a single-board, beam-forming setup
• Design and implement algorithms for beam forming
  with RFI rejection.
• Analogue Beam-former solutions
• ASTRON Time or phase modulation circuit design
  and other beamforming components in GaAs
  technology
• OPAR Time or phase modulation circuit design and
  other beamforming components in Silicon
  technology
• Demonstrate connections and outputs on optical
  fibre using commercial standards.
• Secondary Objectives
• Evaluation report on the role of photonics in
  beam forming.

2
Analogue Beam Forming

• OPAR low cost analogue beamformer in Silicon. (J
  Pezzani)
• 0.25 µm SiGe technology from Philips
  Semiconductors
• good trade-off between price and performance
• 3 bits phase shifters - 360 phase variation with
  45 step
• gain changed with 0.6 dB step on a 3 dB range
  with a 3 bits control
• A full digital serial interface is built on the
  chip
• Manufacturing of the chip and measurements will
  be made in the next months.
• Results from these measurements will be used for
  a second iteration

Layout of 4 channel 2 beams first prototype
beamformer chip
3
Analogue Beam Forming

• Analogue Beam Former in GaAs (ASTRON)
• Beam forming design for EMBRACE
• See DS5 report (Parbhu Patel/Dion Kant)

4
Photonic Beamforming

• Photonic integrated circuit (PIC) beamformer
  (ASTRON) (P Maat)
• Broadband, ns range, true time delay -
  (integrated) optical ring resonators
• Continuously tunable effective delay controlled
  by tuning coupling coefficients between the
  waveguide and the ring resonator, and the
  round-trip phase shifts of the ring resonator.
• First goal develop and study a tile beamformer
  system using PIC beamformer component.
• Next, the photonic beamfomer will be applied in a
  tile demonstrator (EMBRACE).
• Work so far
• specification of the requirements for the
  photonic beamformer.
• development of a low cost, high performance
  optical analogue link
• First photonic beamformer chip samples will
  become available in the autumn of 2006.

5
Digital Beam Forming

• Digital beam forming also part of EMBRACE (DS5)
  (LOFAR)
• Beam forming for All-Digital Aperture Array (M
  Jones Oxford, A Faulkner
  Manchester, P Alexander Cambridge)
• 2-PAD demonstrator
• Can we build a practical (affordable!)
  all-digital system?
• Try to design architecture for flexible beam
  forming system Moores law will catch up
• Use digital system to correct analogue errors
  may be essential for dynamic range spec (eg polzn
  purity)
• Multiple fields of view, up to all sky may be
  essential for dynamic range spec (bright sources)

6
Tile Processor
4-bit, 1.2GS/s, Element Data
Horiz. Polarisation
Vert. Polarisation
Freq. Split. n Chips
210 Polyphase filter
210 8-bit Preset coefficients
210 Polyphase filter
210 8-bit Preset coefficients
X
X
1 pair of 256 elements


210 spectral channels 0 4-bit 1.2MS/s
1023
0
1023
Linear Matrix Mult. to correct Element
polarisation at specific freq. (1 of 210)
..
..
0
4-bit data
255
0
255 1.2MS/s
Inter-element scaling (matrix mults)
Inter-element scaling (matrix mults)
..
..
0
0
255 1.2MS/s
255
16 x 16 element, 2-D FFT (1 of 210) Horiz. pol
16 x 16 element, 2-D FFT (1 of 210) Vert. pol
..
..
FOVs
255 1.2MS/s
8-bit data 0
256 ? 8 FOV selector (1 of 210)
256 ? 8 FOV selector (1 of 210)
7
0
7
8-bit data 0
Linear Matrix Mult. to correct Field-of-View
polarisation at specific freq. (1 of 8)
..
..
255 1.2MS/s
210/m ? 1 Data multiplexer (1 of 8)
210/m ? 1 Data multiplexer (1 of 8)
2-D FFT, m Chips
Field of View 1 of 8 Dual Polarisation, 8 bit
1.2/m GS/s
7
Tile Processor
4-bit, 1.2GS/s, Element Data
Horiz. Polarisation
Vert. Polarisation
Freq. Split. n Chips
210 Polyphase filter
210 8-bit Preset coefficients
210 Polyphase filter
210 8-bit Preset coefficients
X
X
1 pair of 256 elements
Multiplexed data from elements brought onto
frequency splitter chips both polarisations
from an element always on the same Freq. Splitter
chip.


210 spectral channels 0 4-bit 1.2MS/s
1023
0
1023
Each polarisation for each element is split into
1024 spectral bands, corrected for amplitude and
phase if necessary. Total 256 pairs of polyphase
filters
Linear Matrix Mult. to correct Element
polarisation at specific freq. (1 of 210)
Each element at each spectral channel has its
polarisation purity optimised at ONE selected
scan angle.
..
..
0
4-bit data
255
0
255 1.2MS/s
Inter-element scaling (matrix mults)
Inter-element scaling (matrix mults)
..
..
0
0
255 1.2MS/s
255
The corrected polarisation data as spectral
channels are multiplexed as 256/n for each link
to the 2-D FFT chips.
16 x 16 element, 2-D FFT (1 of 210) Horiz. pol
16 x 16 element, 2-D FFT (1 of 210) Vert. pol
..
..
FOVs
255 1.2MS/s
8-bit data 0
256 ? 8 FOV selector (1 of 210)
256 ? 8 FOV selector (1 of 210)
7
0
7
8-bit data 0
Linear Matrix Mult. to correct Field-of-View
polarisation at specific freq. (1 of 8)
..
..
255 1.2MS/s
210/m ? 1 Data multiplexer (1 of 8)
210/m ? 1 Data multiplexer (1 of 8)
2-D FFT, m Chips
Field of View 1 of 8 Dual Polarisation, 8 bit
1.2/m GS/s
8
Tile Processor
4-bit, 1.2GS/s, Element Data
Data from ALL 256 elements over a chosen
frequency range is brought onto one 2-D FFT chip
Horiz. Polarisation
Vert. Polarisation
Freq. Split. n Chips
210 Polyphase filter
210 8-bit Preset coefficients
210 Polyphase filter
210 8-bit Preset coefficients
X
X
1 pair of 256 elements


210 spectral channels 0 4-bit 1.2MS/s
1023
0
1023
Opportunity to correct any Tile level variations
and mutual coupling between elements at each
frequency
Linear Matrix Mult. to correct Element
polarisation at specific freq. (1 of 210)
Form ALL 256 FOVs (beams) over each spectral
channel
..
..
0
4-bit data
255
0
255 1.2MS/s
Inter-element scaling (matrix mults)
Inter-element scaling (matrix mults)
..
..
Select 8 FOVs from possible 256. This is done
with each spectral channel
0
0
255 1.2MS/s
255
Further refining polarisation purity for the
selected FOVs. This can be done specifically for
all 8 FOVs at each spectral channel
16 x 16 element, 2-D FFT (1 of 210) Horiz. pol
16 x 16 element, 2-D FFT (1 of 210) Vert. pol
..
..
FOVs
255 1.2MS/s
8-bit data 0
256 ? 8 FOV selector (1 of 210)
256 ? 8 FOV selector (1 of 210)
Multiplex all the spectral channels processed on
the 2-D FFT chip for transmission to FOV processor
7
0
7
8-bit data 0
Linear Matrix Mult. to correct Field-of-View
polarisation at specific freq. (1 of 8)
..
..
255 1.2MS/s
210/m ? 1 Data multiplexer (1 of 8)
210/m ? 1 Data multiplexer (1 of 8)
2-D FFT, m Chips
Field of View 1 of 8 Dual Polarisation, 8 bit
1.2/m GS/s
9
Key points

• Initial frequency split ? channel bandwidth low ?
  get many elements on one processing chip
• FFT beamforming gives you the whole sky (put eg
  many beams on bright sources for calibration)
• Trade beam area/bandwidth at constant total data
  rate
• Need serious processing chip
• Chip developers say processing is OK on SKA
  timescale
• Problem is I/O! - on chip scale as well as
  station-correlator

10
Future

• Rest of 2006
• Continue paper designs/blue sky thinking
• Recruit hardware and algorithm development team
• 2007
• Prototype key hardware elements
• Algorithm development
• 2008
• Q1 2-PAD final design
• Build 2-PAD beam forming
• 2009
• Results and reports