Download Implementation of an Antenna Array Signal Processing Breadboard

Survey
yes no Was this document useful for you?
   Thank you for your participation!

* Your assessment is very important for improving the workof artificial intelligence, which forms the content of this project

Document related concepts

GPS signals wikipedia , lookup

Radio direction finder wikipedia , lookup

CanSat wikipedia , lookup

Transcript
National Aeronautics and Space Administration
Jet Propulsion Laboratory
California Institute of Technology
Implementation of an Antenna Array Signal Processing
Breadboard for the Deep Space Network
RadioNet Engineering Forum
Workshop:
Next Generation Correlators for Radio
Astronomy and Geodesy
Robert Navarro
Jet Propulsion Laboratory,
California Institute of Technology.
June 27, 2006
RN - 1
03/08/05
National Aeronautics and Space Administration
Jet Propulsion Laboratory
California Institute of Technology
Implementation of an Antenna Array Signal Processing
Breadboard for the Deep Space Network
DSN Large Array Background
Large Array Overview
•
•
•
The DSN is currently evaluating the use
of a Large Array to replace or augment
the current 34 and 70 meter antenna
assets.
Will mainly be used to support telemetry
of deep space missions but will also
support navigation and some science
requirements.
The current Large Array strawman design
calls for
– Three Complexes (Western US,
Australia, European Longitude),
each with:
• 100 to 400 downlink antennas
nominally sized at 12m each.
• Remotely conduct all real-time
monitor and control for the
network
Proposed Signal Processing Req.
•
•
•
•
•
•
Number of Antennas in a cluster scaleable up
to 400.
– Design needs to accommodate growth
but is not infinitely scalable
Number of IF inputs per antenna ~2.
– RCP & LCP or X and Ka
IF signal bandwidth ~500 MHz 1dB
– Main driver for sample rate of 1280
Ms/sec.
Signals of interest can come from anywhere
in input passband
Provide up to 16 simultaneous phased array
outputs
Provide a wideband correlator (~500 MHz)
which can process a significant number of the
antenna signals
– Required to support the array for phase
and antenna position calibration and
searching for lost spacecraft
RN - 2
03/08/05
National Aeronautics and Space Administration
Jet Propulsion Laboratory
California Institute of Technology
Implementation of an Antenna Array Signal Processing
Breadboard for the Deep Space Network
Breadboard Array Signal Processing Objectives
• Provide a means to evaluate the performance of the Breadboard Array’s antenna
subsystem (the antenna, feed and RF-IF downconversion) for one 12 meter and two 6
meter antennas.
• Design and build prototype signal processing hardware
– High density IF digitizer, current design supports 6 Antennas (12 IF inputs) in one
chassis
– High speed FPGA digital signal processing board for AdvancedTCA chassis
• Demonstrate and evaluate proposed signal processing techniques
– Implement an architecture similar to an FX correlator but includes synthesis processing at
the output to reconstruct a time domain signal for beamforming applications
– Provide both beamformer and wideband correlation functions
– Support Order N and Order N2 complexity combining algorithms
– Polyphase FIR filter and FFT for Analysis filterbank
– Synthesis filterbank for reconstruction of beamformer output
• Gain experience with various technologies that may be used in the Large Array
– High speed serial digital signal interconnects (3 to 10 Gbit/s links)
– High speed analog to digital converters (1280 Ms/s, 8 bit)
– AdvancedTCA Chassis with high speed serial backplane
– Field Programmable Gate Arrays for digital signal processing
– Linux OS for use in embedded processors
RN - 3
03/08/05
National Aeronautics and Space Administration
Jet Propulsion Laboratory
California Institute of Technology
Implementation of an Antenna Array Signal Processing
Breadboard for the Deep Space Network
Breadboard Hardware
Breadboard Array Signal Processing
IF Digitizer (IFD)
IFA
IFB
Array Processing Element (APE)
IFA
IFB
Array Processing Element (APE)
Array Sampler
Module (ASM)
Array Sampler
Module (ASM)
Array Processing Element (APE)
IFA
100 MHZ
Common
Module (CMM)
Backplane
IFB
Array Sampler
Module (ASM)
Backplane
700-1200 MHz
Analog Fiber Optic
From Antenna
Realtime Signal Proc (RSP)
12x Digital
Fiber Optic
Timing Reference
1PPS
Gigabit
Etherswitch
Data Processor &
Controller PC (DPC)
RN - 4
03/08/05
National Aeronautics and Space Administration
Jet Propulsion Laboratory
California Institute of Technology
Implementation of an Antenna Array Signal Processing
Breadboard for the Deep Space Network
Breadboard Signal Processing Flow
Breadboard Array Correlator and Combiner Signal Flow
Simplified Two Antenna Case
IF signal
from Antenna
IF signal
from Antenna
ADC
Delay
Analysis
Filterbank
ADC
Delay
Analysis
Filterbank
N
N
Phase
Rotation
Corr
Phase
Rotation
Combine
Cross
Correlation
Data
Synth
Filterbank
Receiver
• Functional Blocks in green are implemented.
• Functional Blocks in blue are in development
RN - 5
03/08/05
National Aeronautics and Space Administration
Jet Propulsion Laboratory
California Institute of Technology
Implementation of an Antenna Array Signal Processing
Breadboard for the Deep Space Network
Detail – Analysis Filterbank
N – Sample Shift Register
(Direction of Shift)
•
•
•
•
Block Diagram of Analysis
Filterbank in Antenna FPGA
This structure allows for
oversampling of the DFT to obtain
overlap in the bandwidth of the
frequency channels.
The extra bandwidth in each
frequency channel allows for fringe
rate (frequency) corrections to be
made to each channel.
Also, it allows for the
analysis/synthesis filterbank
combination to give near perfect
reconstruction with a simple
square root raised cosine prototype
filter.
Input x(n) in Blocks
of M samples
h(  r )
Analysis Window
Windowed
Sequence
+
Time aliasing
(Overlap-Add) in
blocks of K samples.
+
+
Time Aliased sequence
r=0
r=K-1
DFT
~
X
k
(m)
Implemented with FFT
Short Time Transform
Sequence of Operations for weighted overlap-add spectral analyzer
(From Multirate Digital Signal Processing, Crochiere and Rabiner, page 318)
RN - 6
03/08/05
National Aeronautics and Space Administration
Jet Propulsion Laboratory
California Institute of Technology
Implementation of an Antenna Array Signal Processing
Breadboard for the Deep Space Network
APE Board in Real-time Signal Processor
Array Processing Element (APE)
Digital IF
Data From Sampler
1280 Ms/sec
Digital IF
Data From Sampler
1280 Ms/sec
Gigabit
Ethernet
Demux
Demux
PPC
Interface
FPGA
Fanout &
IF Select
FPGA
Ethernet
to/from DPC
PPC
PPC Bus
DSP
FPGA
DSP
FPGA
DSP
FPGA
DSP
FPGA
4
4
High-speed(HS) serial IO
4
4
Crossbar
Switch
4
HS Serial IO to/from
ATCA Fullmesh
Backplane
RN - 7
03/08/05
National Aeronautics and Space Administration
Jet Propulsion Laboratory
California Institute of Technology
Implementation of an Antenna Array Signal Processing
Breadboard for the Deep Space Network
APE Board Picture
RN - 8
03/08/05
National Aeronautics and Space Administration
Jet Propulsion Laboratory
California Institute of Technology
Implementation of an Antenna Array Signal Processing
Breadboard for the Deep Space Network
BBA Signal Processing Results Summary
•
•
•
Built and Tested Hardware: 3 Array Sampler Modules (ASM) for the IF
Digitizer (IFD) and 3 Array Processing Element (APE) boards for the Real
Time Signal Processor (RSP).
Software and Signal Processing Firmware (FPGA code) for the
wideband correlator completed and tested.
Successfully demonstrated these technologies for BBA:
– High Speed A/D (1280 Ms/sec, 10 bit digitizer)
– Analog Fiber link from Antenna to IFD.
– Optical Fiber link from IFD to RSP.
– High Speed Serial links (3.2 Gb/sec) between Filterbanks and
Correlator blocks.
– Real-time Linux OS for Embedded PowerPC processors.
– Wideband (640 MHz) Discrete Fourier Transform Analysis Filterbank
implemented in FPGA.
RN - 9
03/08/05
National Aeronautics and Space Administration
Jet Propulsion Laboratory
California Institute of Technology
Implementation of an Antenna Array Signal Processing
Breadboard for the Deep Space Network
Experimental Results
•
•
•
•
•
Successfully detected interferometric fringes from two 6 meter antennas
using Venus as the source in Dec 2005.
Successfully stopped interferometric fringes using Geometric models
while viewing Venus, Cygnus A, and Cassiopeia A with the two 6 meter
antennas on the mesa in January 2006.
Signal Processing Monitor plots give visibility to confirm correlation
and measure delay offsets but not to do detailed analysis of data.
Correlation Data is archived for later processing using AIPS software to
determine more accurate antenna position.
Continued observations with other sources such as 3C84, 3C273, and
3C48.
RN - 10
03/08/05
National Aeronautics and Space Administration
Jet Propulsion Laboratory
California Institute of Technology
Implementation of an Antenna Array Signal Processing
Breadboard for the Deep Space Network
Experiment Results: Phase and Amplitude
Results from two 6 meter
antennas looking at Cygnus A:
• Amplitude and Phase shown
over 640 MHz complex sampling
band.
• Large spikes in Amplitude and
Phase from RF interference at
X-band
•Geometric models and offsets
for path delay applied to bring
delay to zero.
• For zero delay, the plot of
phase versus frequency should
have a slope of zero across the
band.
RN - 11
03/08/05
National Aeronautics and Space Administration
Jet Propulsion Laboratory
California Institute of Technology
Implementation of an Antenna Array Signal Processing
Breadboard for the Deep Space Network
Experiment Results: Lag and Time Delay Plots
Results from two 6 meter
antennas looking at Cygnus A:
• Plot of Lag Amplitude made
by taking inverse FFT of
frequency channel data.
•Main Lobe of lag plot centered
at delay of Antenna2 to
Antenna1.
• Frequency channels with RF
interference excluded in
calculating these plots.
• Delay Time History plot tracks
peak of Lag Amplitude plot.
RN - 12
03/08/05
National Aeronautics and Space Administration
Jet Propulsion Laboratory
California Institute of Technology
Implementation of an Antenna Array Signal Processing
Breadboard for the Deep Space Network
Experimental Results: Visibility Phase and Amp
3C84 at X-band - day 06/159
Visibility Amplitude (dB)
0
19:00:00
-5
20:00:00
21:00:00
22:00:00
-10
23:00:00
4
0:00:00
3
2
-15
-20
1
-25
0
-30
-1
-35
-2
-40
-45
-3
-50
-4
Visibility Phase (radians)
Lag(0) data for 160 channels (9100 to 9300 MHz)
Running Boresight Script to go on and off source.
Peak_Amp
Peak_Phase
Time UT Hours
RN - 13
03/08/05
National Aeronautics and Space Administration
Jet Propulsion Laboratory
California Institute of Technology
Implementation of an Antenna Array Signal Processing
Breadboard for the Deep Space Network
Scaling to Larger Arrays
•
•
•
Revise current ATCA APE board into 3 or 4 ATCA boards optimized to the various
functions of the Array Signal Processing using latest FPGA technology available.
– Input analysis filterbank (Frequency Channelizer) Board
– Data Routing/Corner Turner Board
– Correlate/Combine Board
– Synthesis Board
Research interconnect strategies for Advanced TCA to Advanced TCA chassis
interconnection.
Architecture must provide features to increase system reliability
– Redundant power supplies & hard drives to avoid the most common failures
– Hot swap capability to allow hardware repair without shutting down systems
– Avoid single points of failure by distributing functions across multiple boards and
chassis
– Automatic diagnostics to identify hardware to be swapped
– Many of these features are provided by attributes of Advanced TCA shelf
technology.
RN - 14
03/08/05
National Aeronautics and Space Administration
Jet Propulsion Laboratory
California Institute of Technology
Implementation of an Antenna Array Signal Processing
Breadboard for the Deep Space Network
Large Array Signal Processing Architecture
FX Beamformer & Correlator
•
•
•
•
•
•
•
•
Digitize entire antenna IF bandwidth
Apply course delay and phase
corrections in time domain
Analysis filterbank uses polyphase
FIR filter and a FFT to break the time
domain signal up into evenly spaced
frequency channels
Apply fine delay and phase
corrections in the frequency domain
Multiple correction profiles required
to support multiple beams per
antenna
Multiple resolutions required to
support wideband correlation and
spacecraft signals
Pre-Router receives frequency
channel data from multiple antennas
and rearranges the data so that each
output carries data for all input
antennas over a subset of the
frequency channels
Failure of one block does not cause
failure of entire signal processing
system
Ant IF
IF Digitizer
Analysis
Filterbank
Ant IF
IF Digitizer
Analysis
Filterbank
Ant IF
IF Digitizer
Analysis
Filterbank
Ant IF
IF Digitizer
Analysis
Filterbank
Ant IF
IF Digitizer
Analysis
Filterbank
Ant IF
IF Digitizer
Analysis
Filterbank
Pre
Router
Pre
Router
Pre
Router
RN - 15
03/08/05
National Aeronautics and Space Administration
Jet Propulsion Laboratory
California Institute of Technology
Implementation of an Antenna Array Signal Processing
Breadboard for the Deep Space Network
Large Array Signal Processing Architecture
FX Beamformer & Correlator
•
•
•
•
•
•
Post-Router receives input from each
Pre-Router and completes the corner
turning process
Beamformers receive data from PostRouters and process data from all
antennas for a subset of the
frequency channels
Beamformers provides both a
wideband correlation function and
narrower band spacecraft processing
Synthesis filterbanks receive data
from multiple beamformers and
transforms them into a wider
bandwidth time-domain signal
Multiple Synthesis filterbanks provide
multiple phased array outputs
Output rate of Synthesis filterbanks
can be scaled to meet required
output bandwidth
Post
Router
Beamformer
Correlator
Synthesis
Filterbank
Post
Router
Beamformer
Correlator
Synthesis
Filterbank
Post
Router
Beamformer
Correlator
Synthesis
Filterbank
Beam 1
Beam 2
Beam K
RN - 16
03/08/05
National Aeronautics and Space Administration
Jet Propulsion Laboratory
California Institute of Technology
Implementation of an Antenna Array Signal Processing
Breadboard for the Deep Space Network
Large Array Signal Processing Architecture
Signal Processing Hardware for strawman 400 Antenna Array
23 Digitizer Racks
1 Antenna (2 bands)
per Sampler
module
6 Sampler & 1
Common module
per chassis
67 Dig chassis per
400 Antenna
Cluster
Each Dig chassis
requires
100MHz, 1PPS,
& 10Base-T
10 Antenna
Signal Proc Racks
8 Beam Former
Signal Proc Racks
2 Antenna (2 bands)
per Ant board
10 Ant & 4 Pre-Rtr
boards per
chassis
20 Ant chassis per
400 Antenna
Cluster
Each Ant chassis
requires
100MHz, &
1000Base-T
1 Beam per Beam
Former chassis
5 Post-Rtr, 8 Beam
Former & 1
Synthesis board
per BF chassis
16 Beamformer
chassis per 400
Antenna Cluster
Each BF chassis
requires
100MHz, &
1000Base-T
RN - 17
03/08/05
National Aeronautics and Space Administration
Jet Propulsion Laboratory
California Institute of Technology
Implementation of an Antenna Array Signal Processing
Breadboard for the Deep Space Network
BBA Correlator Summary
•
•
•
•
•
•
•
•
•
•
•
•
Type: Connected Element
Output: Spectral Line, Continuum, and Combiner
Input BW and Digitization: Digitize at 1280 Msamples/sec with 8 bit samples. 500 MHz
effective bandwidth.
Special Processing: Feedback for Combiner processing. Manual RFI Signal processing to
identify RFI at X-Band (~8100 MHZ)
Spectral Channels: 512 Spectral Channels. Each channel is has 1.25 MHz bandwidth. Each
channel is sampled at 1.5625 MHz.
Integration time: Range of 10 millisecond to 50 seconds.
Dynamic Range: Input A/D has 8 bits dynamic range. Correlation of spectral channels for
phasing and radio astronomy uses 2 bit complex data for each channel. Combiner output is 6
bits complex per spectral channel and 8 bits real output for synthesized time domain signal.
Scale of Construction: Pushing technology limits
Technology approach: Custom Hardware in ATCA chassis. FPGA’s with 2.5 Gigasample/sec
serial links, Embedded Real-time Linux
Current hardware scalable up to 16 antennas per ATCA chassis. Future enhancements to
architecture should enable scaling to 100’s of antennas.
The architecture is very flexible. The signal processing board contains 4 signal processing
FPGA’s with over 50000 logic cells, 232 18bit by 18 bit multipliers and 16 high speed serial
I/O’s each. All FPGA’s interconnected to each other and across the ATCA backplane.
Currently, the main limiting factor is the amount of data transported over the 2.5 Gs/sec serial
IO links.
Correlator uses an FX architecture with polyphase filtering and independent fringe phase
correction for each channel.
RN - 18
03/08/05