1
National Radio Astronomy Observatory
The EVLA --
Status, Future
Rick Perley
National Radio Astronomy Observatory
2
EVLA Project Overview
• The Expanded Very Large Array is a major upgrade of the Very Large Array.
• The fundamental goal is to improve all the observational capabilities of the VLA -- except spatial resolution -- by at least an order of magnitude
• Counting all sources, a $90M project. • The Project began in 2001, and will be
completed in 2012 – on time, on spec, on budget.
3
Key EVLA Project Goals• Full frequency coverage from 1 to 50 GHz.
– Provided by 8 frequency bands with cryogenic receivers.• Up to 8 GHz/polarization instantaneous bandwidth
– Provided by two independent dual-polarization IF pairs, each of up to 4 GHz bandwidth per polarization.
• New correlator with 8 GHz/polarization capability– Designed, funded, and constructed by our Canadian
partners, HIA/DRAO– Essentially comprises 64 separate full-polarization ‘sub-
correlators’, each of maximum 128 MHz BW with 256 channels – a total of 16384 channels.
– Unprecedented flexibility in allocating resources. • <1 mJy point-source continuum sensitivity (1-s, 12-
Hr), (2 mJy at L and Q bands)• Noise-limited, full-field imaging in all Stokes
parameters for most observational fields.
4
Overall EVLA Performance Goals
Parameter VLA EVLA Factor
Point Source Sensitivity (1-s, 12 hr.) 10 mJy 1 mJy 10
Maximum BW in each polarization 0.1 GHz 8 GHz 80
# of frequency channels at max. BW 16 16,384 1024
Maximum number of freq. channels 512 4,194,304 8192
Coarsest frequency resolution 50 MHz 2 MHz 25
Finest frequency resolution 381 Hz 0.12 Hz 3180
# of full-polarization sub-correlators 2 64 32
(Log) Frequency Coverage (1 – 50 GHz) 22% 100% 5
The EVLA’s performance will be vastly better than the VLA’s:
5
The EVLA as a ‘Leveraged Investment’
• The Project is making maximum use of existing hardware and infrastructure.
• The EVLA is recycling the VLA’s 25-meter parabaloids.
– Off-axis Cassegrain optics. – Change band by rotating
subreflector• Disadvantages:
– Big and slow – Not the best choice for fast
wide-field surveys• Advantages:
– Paid for!– Work well – even up to 50
GHz.• Also retained:
–Configurations, buildings, basic infrastructure, people.
Antenna 24 – the first EVLA antenna outfitted with all eight feeds.
6
Full Frequency Coverage with Outstanding Performance
• There are eight feeds, tightly packed around the secondary focus feed ring.
Band (GHz)
Tsys/e(best weather)
1-2 L (60)
2-4 S 55 -- 70
4-8 C 45 -- 60
8-12 X 45*
12-18 Ku 50*
18-26.5 K 70 -- 80
26.5-40 Ka 90 -- 130
40-50 Q 160 - 360
L
K
QKa
XC
S
Ku
FeedHeaters
* : Anticipated value
Antenna Conversion Progress
• VLA antennas go into the ‘barn’ to be converted to EVLA standards.– New LO/IF system– New receivers, feeds, etc. – Digitizers, FO, power, etc.
• Status – nearly completed:– 23 antennas fully converted, and back in operation.– 1 undergoing conversion– 4 to go – process completed by mid 2010.
7
Receiver Implementation Schedule
• Receiver implementation/upgrade schedule is quite complicated. – Two bands (40 – 50, 18 -- 26 GHz) are essentially unchanged
• These are deployed on converted antennas directly.– Two bands (2 – 4, 26 – 40 GHz) are brand new.
• Design complete, implementation underway in field.– Remaining four bands are completely redesigned
• New designs implemented on array as they are completed. • In the meantime, ‘interim’ receivers outfitted upon conversion,
upgraded to new design in field.
• Receiver outfitting pace determined by funding and manpower. – Will be completed by end of 2012.
8
9
Full-Band Receiver Availability Timescale• Note that during transition, L, C, X, K and Q band receivers are on
all antennas.
VLA Correlator WIDAR Correlator
The ‘WIDAR’ Correlator
• The key element of the EVLA is its Canadian funded `WIDAR’ correlator – a 10 petaflop computer.
• Designed and built by the correlator group at HIA/DRAO in Penticton BC, to meet or exceed NRAO requirements.
• WIDAR is a ‘full-service’ correlator, designed to meet the diverse needs of our user community.
• Major capabilities:– 8 GHz/polarization maximum instantaneous bandwidth– Full polarization– Minimum # channels – 16384, up to 4.2 million maximum. – Spectral dynamic range up to 58 dB– Extensive special modes– 64 independently tunable sub-band pairs, each pair effectively
forms an independent ‘sub-correlator’.10
11
The ‘WIDAR’ Correlator• Accepts four input ‘baseband’ pairs of up to 2 GHz BW each.• Each input is digitally sub-divided into 16 ‘sub-band’ pairs.• Each of the 4*16 = 64 sub-band pairs can be considered as a
separate full-polarization ‘sub-correlator’, with 256 channels.
Sub-band
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16BBP
1
2
3
4
R - 2 GHz
L - 2 GHz
R - 2 GHz
L - 2 GHz
R - 2 GHz
L - 2 GHz
R - 2 GHz
L - 2 GHz
1 sub-band pair with RQ = 4:
256 channels for1, 2, or 4 polarization products
Maximum SBW = 128 MHz
Quadrant
Q1
Q2
Q3
Q4
Four Active Baseband PairsInitial Quantization = 3 bits
Re-Quantization = 4 bits4 Input
Baseband Pairs:3-bits @
4.096 Gsamp/sec
Frequency 1
Frequency 2
Frequency 3
Frequency 4
There are 64 independent sub-band pairs, each with its own center frequency, bandwidth, and
polarization combination
12
WIDAR Flexibility
• Each of the 64 sub-band pairs:– Is independently tunable to any frequency*– Has a sub-band width of any of 128, 64, 32, …, .031 MHz– Has recirculation available (doubles the number of spectral
channels for each halving of the sub-band width)– Can utilize the computational resources of any number of other
sub-band pairs (trading bandwidth for channels).
• Number of channels is from 16384 to 4194304• Frequency resolution is from 2 MHz to 0.19 Hz.• Numerous special modes and capabilities:
– Pulsar binning and gating – Phased array– Burst mode– Multiple subarrays
* There are some limitations
• 4 or 7-bit internal mode • 1,2,3,4 or 8 bit input• Online RFI excision• Fully VLBI ready
WIDAR’s Special Modes• Considerable effort went into including special modes,
including:– Pulsar gating and binning:
• 2 banks of 1000 bins each, with 100 msec time resolution.• More bins (to 65536), and higher time resolution (15 msec)
available by reducing spectral channels. – Phased Array with the full bandwidth– VLBI capabilities built in.– Multiple subarrays (up to 8)– Burst mode
• 100 msec time resolution with 65K channels• Faster possible with larger CBE
• Basic observing modes will be commissioned first. • Special modes come later.
13
14
Results from Prototype Correlator Testing
• 1244-1756 MHz• 8192 x 62.5 kHz
(13 km/s for local HI)
• Final EVLA:–512 MHz (z=0-0.3)–@ 7.8 kHz (1.7
km/s)
HIVLApolarizer
satellites
ABQradars
Current VLA: 6.25 MHz @ 98 kHz512 MHz
• 12-minute Vector average of a single short baseline, on 3C84.
• Eight sub-bands, each 64 MHz wide, each with 1024 channels.
Dec
ibel
Pow
er D
ensi
tyP
hase
WIDAR Installation• The WIDAR correlator is now being installed into its specially
shielded room at the VLA site. • All racks, power, cabling, FO are completed. • Initial WIDAR-0 now operating – for testing only.• Completion of board installation early in 2010. • All 128 Station Boards and 128 Baseline Boards will be on site by
March 2010.
Air ConditionersKen Sowinski
(6’0”) Power Backup/Conditioners~100 dB Shielded Chamber
WIDAR-0
• This is the name given to the initial configuration of the final correlator.
• Current setup:– 12 antennas, one IF dual-polarization pair. – Full polarimetry enabled– Four subbands, each of 256 channels/correlation.
• Testing is currently focused on system integration, reliability, and stability.
• Once these are established, precision imaging tests are critical to establishing full system performance.
16
L-Band Spectrum
17
• Four subbands, covering 512 MHz, with 500 kHz resolution.• With 128 subbands total to edit, these illustrate the challenge in
editing.• Most of the notable RFI has well understood origins.
1300 1428 1556 MHz 1684 1812
Long spacing~3 Km
Short spacing~0.1 Km
GPSL3
FAA Radars
Inm
ars
at
GPSL1
Glo
na
ss
Irid
ium Meteorology
L-Band ‘Dead Zone’
3C286 Deep Field
• Detailed testing underway with 12-station, full polarization, 4 subband initial configuration.
• Most demanding testing is at L-band, short spacings.
• Best image so far: 200,000:1 DR for 3C286.
• Some artifacts visible
18
19
Initial WIDAR Capabilities for Science
• VLA correlator will be decommissioned on Jan 11, 2010.• The array will return to service, using WIDAR, around
March 1, with all converted EVLA antennas (26).• Two basic initial observational modes:
1. Two independently tunable sub-band pairs with 64 channels for each of four cross-correlations (total 512 channels), OR
2. One tunable sub-band pair, with 256 channels for each of two cross-corrrelations (512 total channels).
– For both, sub-band width adjustable to one of 128/2n MHz (n = 0, … 12).
• Growth in observational capabilities will be set by growth of software capabilities.
OSRO and RSRO
• OSRO = ‘Open Shared Risk Observing’. – Is in fact ‘business as usual’. – Observers will access EVLA in same manner as current for VLA.
– For continuity with existing VLA observing, the initial WIDAR
capabilities have been based on existing VLA capabilities. • But in fact, observers will get much, much more.
• RSRO = ‘Resident Shared Risk Observing’– For those willing to spend significant time in Socorro, and have
skills & interest in assisting in implementing advanced correlator modes or calibration methodologies.
– Participants will have much more extensive capabilities made available to them.
– Participants with the right skills should have the opportunity to greatly accelerate the development/availability of special observing modes. 20
Initial ‘OSRO’ Capabilities• By March 2010, we plan to provide the following for
‘normal, external’ observers.– Table 1: Full Polarization Continuum Applications
21
Sub-BW (MHz)
Sub-BW (MHz)
# Chan/ product
Chan Width (kHz)
Velocity Width (km/sec)
Velocity Coverage (km/sec)
128 4 64 2000 600/nG 38400/nG
64 4 64 1000 300/nG 19200/nG
32 4 64 500 150/nG 9600/nG
… 4 64 … … …
0.0625 4 64 0.98 0.29/nG 18.75/nG
0.03125 4 64 0.488 0.15/nG 9.375/nG
There will be two independently tunable sub-bands, each with the capabilities noted in the table.
Initial OSRO Capabilities (cont.)
Table 2: Spectral Line Applications
22
Sub-BW (MHz)
Sub-BW (MHz)
# Chan/ product
Chan Width (kHz)
Velocity Width (km/sec)
Velocity Coverage (km/sec)
128 2 256 500 150/nG 38400/nG
64 2 256 250 75/nG 19200/nG
32 2 256 124 37.5/nG 9600/nG
… 2 256 … … …
0.0625 2 256 0.244 .073/nG 18.75/nG
0.03125 2 256 0.122 0.037/nG 9.375/nG
For both modes: • minimum integration time is 1 second.• Doppler tracking will be available.• Data can be reduced via AIPS (but CASA is encouraged!)
The RSRO Program
• Program will run from March 2010 to end of 2011.• Application through regular proposal deadlines
– Special rules apply – see website, or contact Claire Chandler at
• Intent is to attract skilled individuals to assist with commissioning process.
• Minimum residency ~ 3 months. Time available proportional to time spent in Socorro.
• Up to 25% of all observing time to be made available. • NRAO should be able to provide housing in Guest House• Applicants with their own support will definitely have
advantage. NRAO support level unknown …• Full details on: www.aoc.nrao.edu/evla/astro/rsro.shtml
23
24
WIDAR Growth
• The expansion in observational capabilities will be rapid, but measured.
• All initial observations will be with the ‘fundamental homogeneous correlator setup’– All sub-band pairs with the same width and channelization,
arranged to maximize total bandwidth coverage. • Resident observers (RSRO program) should have
access to:– 2 GHz/polarization BW (all antennas) by mid-2010– 8 GHz/polarization BW (all antennas) by end of 2011. – Recirculation by late 2010– Independent sub-band tuning by early 2011– Flexible resource allocation by mid 2011.
• Schedule for availability of special modes dependent on availability of resources.
Final WIDAR Capabilities
• The ultimate WIDAR correlator’s capabilities are too vast to describe in one or two slides.
• To give a flavor, I give two examples based on two basic correlator modes:1. The Basic Homogenous Mode – full band coverage
• All subbands arranged to cover the entire BW of a particular band
• All subbands have the same width and number of channels.
2. Targeted sub-bands, with recirculation.• Each sub-band tuned to a specific frequency, with individual
sub-band width and spectral resolution.
25
Summary of Wide-band Coverage:
Freq. IF BW SBW # SBP Dn Dv Nch per Nch
GHz GHz MHz kHz km/s spctrm total
L 1—2 1.024 16 64 125 25 128 16384
S 2—4 2.048 32 64 250 25 128 16384
C 4—8 4.096 64 64 500 25 128 16384
X 8—12 4.096 64 64 500 15 128 16384
U 12—18 6.144 128 48 1000 20 128 12288
K 18—26.5 8.192 128 64 1000 13 128 16384
A 26.5—40 8.192 128 64 1000 9 128 16384
Q 40--50 8.192 128 64 1000 6.5 128 16384
• For dual (RR,LL) polarization, with no recirculation• For full polarization, resolutions are 2 x poorer.
•The data rate is 6.2 MByte/sec for 10-second integration (4.8 at U-band)
•The resulting data volume is 22 GB in 1 hour (17 GB at U-band)
Full-Band Coverage, with Recirculation.
• Recirculation is scheduled for availability in T3 2010. – Recirculation doubles the number of channels for each halving of the
sub-band width• There will be no changes to Q, Ka, K and Ku bands, since the full
128 MHz sub-band width is required to provide full band coverage.
• For dual polarization, capabilities in L, S, C and X bands become:Freq. IF BW SBW #
SBPDn Dv Nch
perNch Rate (10 s avg)
GHz GHz MHz kHz km/s spctrm total MB/sec GB/Hr
L 1—2 1.024 16 64 15.6 3.1 1024 131072 50 178
S 2—4 2.048 32 64 62 6.3 512 65536 25 89
C 4—8 4.096 64 64 250 12.5 256 32767 12 45
X 8—12 4.096 64 64 250 7.5 256 32767 12 45
• For full polarization, channel widths/resolutions are doubled.
Some Science Applications
• Summary of coverage: – With full-band frequency coverage, the entire maximum
bandwidth can be covered with ~10 km/sec velocity resolution. – In general, if higher velocity resolution is required, each factor of
four increase in velocity resolution requires a factor of two reduction in frequency (and velocity) coverage.
• Even this most basic setup enables a huge range of new science capabilities. Some examples:– ~1 mJy/beam sensitivity full-polarization continuum observations
over the full primary beam – all bands. – Wide-band high-redshift surveys of molecules in absorption and
emission.– Deep polarimetric imaging and RM analysis of bright sources,
the galaxy, and clusters.
The Next Step – Flexible Tuning withAdjustable Sub-band Widths.
• Individual tuning of each of the 64 sub-band pairs is scheduled for RSRO availability in T1 2011. – Each of the 64 sub-band pairs would be digitally tunable to any
given frequency within the input bandwidth. – The sub-band width and spectral resolution of each will also be
variable.
• This will enable greatly improved capabilities in studying spectral emission of atomic and molecular emission from specific regions, where:– Full bandwidth coverage is not needed, and/or
– Adjustable spectral resolution is advantageous.
Molecular Line Emission Studies of Massive Star-Forming Regions
• Claire Chandler has proposed two K-band experiments:1. Studies of a Massive Star-Forming Region
– 32 molecular transitions, to be observed at 0.2 km/sec, and– 8 RRLs, to be observed with 1 km/sec, and– some reasonable amount of continuum.
2. Studies of a Cold Dark Cloud. – 54 molecular transitions (mostly heavy molecules) requiring 0.01
km/sec resolution, plus– Some reasonable amount of continuum
• Can WIDAR do this?• Answer: Yes, of course! *
* I wouldn’t give these examples if it couldn’t …
Studies of a Massive star-forming region at K-Band: Goals and Requirements
• Observe high-density tracers NH3, all available transitions from (1,1) to (8,8), and CH3OH; – gives density and temperature structure of hot cores (very
young, massive, protostars)• Observe shock tracers, interaction of protostars with
surrounding cloud: transitions of SO2, H2O, OCS, H2CS, H2CO, OH
• These 32 molecular lines require a velocity resolution of 0.2 km/sec.
• Observe radio recombination lines and continuum emission from a nearby HII region
• The 8 RRL require 1 km/sec velocity resolution. • Observe as much line-free continuum as possible for the
free-free emission
Massive SFR – Correlator Setup• Tune the four available baseband frequency pairs to:
1. 18.6 – 20.6 GHz which covers 3 RRL + 1 Mol (12 SBP free)
2. 20.6 – 22.6 GHz which covers 2 RRL + 3 Mol (11 SBP free)
3. 22.6 – 24.6 GHz which covers 2 RRL + 14 Mol (all SBP used)
4. 24.6 – 26.6 GHz which covers 1 RRL + 14 Mol. (1 SBP free)
• Set the 32 SBPs covering the molecules to a BW = 16 MHz, providing 1024 channels in both RR and LL.
• Set the 8 SBPs covering the RRLs to BW = 32 MHz, providing 512 channels in both RR and LL.
• This leaves 24 SBPs to cover the continuum (at 128 MHz BW each), or for other transitions.
• A total of 79872 channels … a high data rate of 30 MB/sec with 10 sec averaging.
– Could reduce this rate to more manageable size by averaging in time in the CBE (Correlator Backend End).
The Entire Spectrum• Showing the distribution of the SFR lines, color coded by species.• The spans for the four BBPs are as shown.
BBP1 BBP2 BBP3 BBP4
Within BBP #3:
• Showing a ‘close-up’ of the coverage within one of the BBPs.• The green rectangles show the SBP frequency coverage for the
molecules. • The red rectangles shows the (wider) SBP coverage for the RRLs.
1024 channels within green boxes
512 channels within red boxes
Future Challenges
• All this fantastic new capability does not come without challenges.
• Observing Methodologies– Denser calibrator grids, improved referenced pointing
• Archiving: – How to store and distribute datasets generated by
rates exceeding 100 MB/sec (and potentially as high as 1 GB/sec).
• Editing: – Efficiently removing bad data from 256 independent
data streams– Blanking or (better) subtraction of external RFI.
35
Future Challenges (cont.)
• Calibration:– Characterizing, calibrating, and correcting for system
performance variations over 2:1 bandwidth ratios.– Particularly interesting is doing polarization calibration correctly.
• Imaging: – Full beam full Stokes imaging, over 2:1 bandwidth ratio, with
non-coplanar baselines requires efficient implementation of new methodologies.
– Beam corrections in I, Q, U, and V critical. – Eventually will want to combine these with mosaicing.
• Image Analysis:– What to do with all those images? – How to avoid ‘Sensory Overload’?
36
More on Imaging …
• WIDAR will permit, for the first time, non-distorted imaging throughout the entire primary beam.
• This will be enabled by the basic ‘homogenous’ mode. • Imaging the full beam will be relatively straightforwards
at high frequencies (Ku band and higher):– Non-coplanar baseline problem negligible– Very few background sources to remove.
• Inverse situation at L and S bands. – Complete background source removal necessary to reach
thermal noise.
• Software to do full beam, full bandwidth imaging under development.
37
38
Summary
• EVLA will provide a fabulous capability for 1 -- 50 GHz science, with unrivaled sensitivity, resolution and imaging capabilities comparable.
• All antennas converted to new standards by mid 2010.• All receivers completed by end of 2012.• Initial WIDAR capabilities begin March 2010.• 2 GHz/Polarization available by mid 2010.• 8 GHz/Polarization available late in 2011. • Specific observing capabilities and availability timescales
are now being defined• You can influence these by joining the RSRO program!