Äîêóìåíò âçÿò èç êýøà ïîèñêîâîé ìàøèíû. Àäðåñ îðèãèíàëüíîãî äîêóìåíòà : http://www.mrao.cam.ac.uk/projects/skads/lofar/talks/anderson_MKSP_hardware.pdf Äàòà èçìåíåíèÿ: Wed Apr 8 00:23:46 2009 Äàòà èíäåêñèðîâàíèÿ: Tue Aug 18 11:13:58 2009 Êîäèðîâêà: Ïîèñêîâûå ñëîâà: ï ï ï ï ï ï ï ï ï ï ï ï ï ï ï

MKSP Relevant LOFAR Hardware Capabilities
James M Anderson
anderson@mpifr-bonn.mpg.de

On behalf of LOFAR

See also http://www.lofar.org/operations/doku.php?id=science:lofar_oss


LOFAR Magnetism KSP Meeting, Cambridge, 2009 Mar 25 James M Anderson 1/27




Bandwidth
­ ­

Alternate Title: Frequency Coverage is an Orthogonal Concept to Bandwidth for LOFAR
Integration of instrument response as a function of frequency For LOFAR, raw bandwidth: N sb
sb

, effective bandwidth: N ch

ch



Frequency Coverage
­ ­

Location of instrument response as a function of frequency For LOFAR, effective frequency coverage: which subbands/channels are unflagged
Difference between maximum and minimum frequencies



Frequency Span
­

LOFAR Magnetism KSP Meeting, Cambridge, 2009 Mar 25

James M Anderson

2/27


LOFAR : The Low Frequency Array


Aperture array technology
­

digital processing normally 30 to 80 MHz can do 10 to 80 MHz 120 to 240 MHz
­



Low Band (LBA)
­ ­



High Band (HBA)
­



3rd input



Core (2 km diameter) Remote (inside NL) International (outside NL)

}

Orginal LOFAR

}

­

open at International stations extra LBA inputs for Dutch stations (better performance < 30 MHz)

Current LOFAR
3/27

LOFAR Magnetism KSP Meeting, Cambridge, 2009 Mar 25

James M Anderson


C o re


2 km diameter Micky Mouse design Station Beam FWHM
­ ­

8.7 6.6 30

5.3 2.6°

75 120 240 MHz



Synthesized beam
­ ­

800 300 200 100" 30 75 120 240 MHz

LOFAR Magnetism KSP Meeting, Cambridge, 2009 Mar 25

James M Anderson

4/27


Remote


Up to 130 km baselines Circular-pair half-design Station Beam FWHM
­ ­

8.7 6.6 30 20 30

3.7 1.9°

75 120 240 MHz 8 5 3"



Synthesized beam
­ ­

75 120 240 MHz

LOFAR Magnetism KSP Meeting, Cambridge, 2009 Mar 25

James M Anderson

5/27


International


~1000 km baselines Original station design Station Beam FWHM
­ ­

9.9 4.0 30

2.5 1.2°

75 120 240 MHz



Synthesized beam
­ ­

1.7 0.7 0.4 0.2" 30 75 120 240 MHz

LOFAR Magnetism KSP Meeting, Cambridge, 2009 Mar 25

James M Anderson

6/27


Station Electronics: Gory Details 1


Bandpass filter


10--90 MHz 30--90 MHz 110--190 MHz 170--230 MHz 210--270 MHz



12 bit A/D converter
­ ­

200 MHz or 160 MHz clock Forms 100 MHz or 80 MHz bands



RCU modes: common combinations of antenna inputs, bandpass filters, and clock rates assigned special RCU codes
­

But LOFAR can observe with any antenna input, bandpass filter, and clock rate combination


Low band observations with 160 MHz clock potentially important for full frequency coverage for RM synthesis and spectral line work

­

Each antenna/polarization input can chose own antenna and bandpass
James M Anderson 7/27

Must use the same clock frequency LOFAR Magnetism KSP Meeting, Cambridge, 2009 Mar 25


Station Electronics: Gory Details 2


Polyphase filterbank converts the time series data from each dipole/tile into 512 (513) frequency subbands


195.3125 kHz subbands for 200 MHz clock 156.2500 kHz subbands for 160 MHz clock 16 bit complex number (16 bit real, 16 bit imag (200 MHz) or 6.40 s (160 MHz clock)
inary)

for each subband every 5.12 s



Beamformer hardware processes up to 248 of these subbands


248 subbands determine bandwidth available to beamformer


48.4375 MHz for 200 MHz clock 38.7500 MHz for 160 MHz clock

­ ­ ­

Arbitrary subband selection by astronomer Frequency coverage not required to be contiguous Calibration will work best in full production system with wide frequency coverage

LOFAR Magnetism KSP Meeting, Cambridge, 2009 Mar 25

James M Anderson

8/27




16 bit complex samples total bandwidth 48.4375, 38.7500 MHz
­ ­ ­ ­ ­

Station Electronics: Gory Details 3

1 beam 248 subbands (48.4375, 38.7500 MHz for 200, 160 MHz clock) 2 beams 124 subbands each (24.22, 19.38 MHz), may reuse subbands across beams 4 beams 62 subbands each (12.11, 9.69 MHz), may reuse subbands across beams 8 beams 31 subbands each (6.05, 4.84 MHz), may reuse subbands across beams 8 beams, arbitrary subband allocation (248 subbands total, 48.4375, 38.7500 MHz total), may reuse subbands across beams



8 bit complex samples total bandwidth 96.875, 77.500 MHz
­

2 beams of 248 subbands each

­ ­

4 beams o selection, 8 beams o selection,

, all subbands reused f 124 subbands each (24.22, 19.38 MHz), from original 248 subband may reuse subbands across beams f 62 subbands each (12.11, 9.69 MHz), from original 248 subband may reuse subbands across beams
(48.44, 38.75 MHz)



4 bit complex samples total bandwidth 193.75, 155.00 MHz
, all subbands reused ­ 8 beams of 124 subbands each (24.22, 19.38 MHz), from original 248 subband selection, may reuse subbands across beams 3 x 1 GE network bandpass limits observations to ~ 30.5 (61, 122) MHz for 16 (8,4) bit modes, but full bandwidth available for all 10 GE stations
­

4 beams of 248 subbands each

(48.44, 38.75 MHz)



LOFAR Magnetism KSP Meeting, Cambridge, 2009 Mar 25

James M Anderson

9/27


Blue Gene: Gory Details 1


Polyphase filterbank generates 256 channels per subband by default


762.939... Hz channel width for 200 MHz clock 610.352... Hz channel width for 160 MHz clock



Visibility output rate limited to 50 Gb/s by current storage cluster input rate
­

For LOFAR operation (18 Core, 18 Remote, 8 International stations):


LOFAR Core (18 C stations)
­

16, 8, and 4 bit data modes all fit under storage rate limit 16, 8 bit data modes all fit under storage rate limit 4 bit fits for LBA, does not fit for HBA All bit depths have storage rates too high



LOFAR Remote (18 C + 18 R stations)
­ ­



LOFAR International (18 C + 18 R + 8 I stations)
­

LOFAR Magnetism KSP Meeting, Cambridge, 2009 Mar 25

James M Anderson

10/27


Frequency Comb


No requirement to have contiguous frequency coverage Best astronomical performance often achieved with large frequency coverage Distribute subband allocation across the entire band (a frequency comb) Using multiple combs with offset subband selections during the same observation allows the entire frequency range to be covered Can also allocate more subbands around some frequency to weight observations for particular goal
James M Anderson 11/27











LOFAR Magnetism KSP Meeting, Cambridge, 2009 Mar 25


LOFAR

RM

for Cosmic Magnetism


RM error estimate shown as function of polarized signal and observing time Assumed spectral index = -0.7 Observing time distributed to fill HBA frequency range Only get 0.1 rad m-2 in 10 s for a 10 mJy source
­







Need ionospheric Faraday coherence time to be longer than 10 s
12/27

LOFAR Magnetism KSP Meeting, Cambridge, 2009 Mar 25

James M Anderson


User Sensitivity: I


Weather delays in NL pushing back rollout (read 2010 Jan for 2009 Apr) Full (international) LOFAR about 1.7 times better than NL LOFAR HBA sensitivity roughly flat LBA system peaks around 56 MHz Noise increases rapidly toward low frequencies
13/27









LOFAR Magnetism KSP Meeting, Cambridge, 2009 Mar 25

James M Anderson


Long Baselines, RM Synthesis, and Data Volumes
James M Anderson
anderson@mpifr-bonn.mpg.de

On behalf of LOFAR


LOFAR Magnetism KSP Meeting, Cambridge, 2009 Mar 25 James M Anderson 14/27




Station Beam FWHM
­ ­

Why Do We Want International Baselines?
9.9 4.0 30 2.5 1.2° 75 120 240 MHz



Synthesized beam
­ ­

1.7 0.7 0.4 0.2" 30 75 120 240 MHz



Most sensitive stations Small synthesized beam to avoid polarization beamsmearing
James M Anderson 15/27

LOFAR Magnetism KSP Meeting, Cambridge, 2009 Mar 25


Observing the HBA Station Beam


Station beam 2.5 to 1.2 degrees in size (120 to 240 MHz)
­

Assume 2 degree beamsize for imaging Assume typical 4 pixels across beam for imaging



LOFAR synthesized resolution ~ 0.4 to 0.2 arcseconds
­



1.1 x 1010 pixels per station beam (47 GB per frequency) RM resolution 0.1 rad m-2, RM range ±100 rad m-2 2.3 x 1013 pixels per RM cube (93 TB) Extragalactic RM search
­ ­ ­ ­

RM resolution 1 rad m-2, RM range ±10 000 rad m2.3 x 1014 pixels per RM cube (930 TB)

2

RM resolution 0.01 rad m-2, RM range ±10 000 rad m2.3 x 1014 pixels per RM cube (93 PB)
James M Anderson

2

LOFAR Magnetism KSP Meeting, Cambridge, 2009 Mar 25

16/27


Visibility Data Volume

LOFAR Magnetism KSP Meeting, Cambridge, 2009 Mar 25

James M Anderson

17/27


But...


The online storage cluster can only accept 50 Gb/s So the visibility data volume for the MKSP will always be
­

50 Gb/s * t

(6.25 GB/s * t)



100 h observation produces 2.25 PB Can we deal with this?
­

Online storage in Groningen will only have 1 PB, for all LOFAR usage, with data deleted within 1 week of observations



Or for calibration purposes, will we actually need to simultaneously observe 4 or more fields, so that in order to get 100 h on a specific object, we need 400 h total, or 9.0 PB?
­

Where could we put that?



Make RM cubes with small (8 h?) chunks of observations? Average in time and frequency, possibly losing field of view?
James M Anderson 18/27

LOFAR Magnetism KSP Meeting, Cambridge, 2009 Mar 25


RM Range and Frequency Resolution


Default correlator resolution (256 channels per subband) allows one to measure a maximum RM of ±13 000 rad m-2 at 120 MHz |RM| 3 / (2c2ch) 30 200 50 900 120 13 000 240 MHz 100 000 rad m-2 200 MHz clock





For HBA observations, can average significantly in frequency if one is willing to accept a maximum RM of only 1000 or 100 rad m-2

LOFAR Magnetism KSP Meeting, Cambridge, 2009 Mar 25

James M Anderson

19/27


Time and Bandwidth Smearing



1000 km baseline LOFAR, International station beam, 120 MHz Correlator integration times in seconds, frequency averaging in terms of original 256 channels per subband
James M Anderson 20/27

LOFAR Magnetism KSP Meeting, Cambridge, 2009 Mar 25


Smearing Commissioning Tests


Can correct visibilities for time and bandwidth smearing
­ ­ ­

Expensive software operation Small facets? Only works if reduction not too large (0.5? 0.8? 0.95?) Cannot regain lost sensitivity Suppose we need dynamic range of 106 to 107, and we have a software fudge to correct a 10 or 20% smearing loss

­



How well can this be done?
­



What about sources in far sidelobes? How well do we have to subtract them?
James M Anderson 21/27

LOFAR Magnetism KSP Meeting, Cambridge, 2009 Mar 25


MKSP Project Plan, Use Cases, and Commissioning Plan
James M Anderson
anderson@mpifr-bonn.mpg.de

On behalf of LOFAR


LOFAR Magnetism KSP Meeting, Cambridge, 2009 Mar 25 James M Anderson 22/27


Project Plan
A. B. C. D. E. F.


Science background Science goals Source list, or source selection criteria Technical requirements Processing software capabilities General processing requirements

Working group leaders need to write down plans for their working group What can we actually do with LOFAR?



LOFAR Magnetism KSP Meeting, Cambridge, 2009 Mar 25

James M Anderson

23/27


A. Time i. Total time ii. Time distribution a. (u,v) filling b. Over days, weeks, months? B. Number science beams i. Distribution on sky ii. Able to share with other users? C. Number calibration beams D. Visibility details i. integration time ii. Frequency resolution iii. Subband selection E. Switching times i. Beam directions ii. Subbands iii. RCU modes iv. Sources F. Calibrated visibility characteristics i. Integration time ii. Frequency resolution iii. Error levels (RM precision, dynamic range, etc.) G. Image processing i. Images? ii. RM cubes? iii. Resolution (time, frequency, spatial) H. Data storage volume and timescale
LOFAR Magnetism KSP Meeting, Cambridge, 2009 Mar 25

Use Cases

Maybe this should scare you No worse than planning observations for the VLBA or EVN
­

But maybe as a VLBI person my expectations are skewed

Most of you need to learn and understand the instrument properties
­ ­

Theorists excepted Read the LOFAR Observing Status Summary (like you would for the VLA, VLBA, GMRT, ...) Watch the LOFAR Status Meeting every Wednesday
24/27

­

James M Anderson


Calibration Document


Too scary for most of you Ger and myself will write up Others welcome to assist Many people needed to work on commissioning observations, software development, and so on

LOFAR Magnetism KSP Meeting, Cambridge, 2009 Mar 25

James M Anderson

25/27


Commissioning Plan
A. B. C. D. E. F.

­ ­ ­

Tests to do Sources Software Ionosphere People Resources

Need to sort out initial plans during this workshop!!!
Think about this during this afternoon And during the dinner And maybe at the bar after that



And come prepared tomorrow morning to really get to work during the discussion section
­ ­

Tests, sources, time What can we share with other KSPs?
James M Anderson 26/27

LOFAR Magnetism KSP Meeting, Cambridge, 2009 Mar 25


The End


or, I'll shut up for now...

LOFAR Magnetism KSP Meeting, Cambridge, 2009 Mar 25

James M Anderson

27/27