Keith Grainge - Array design

Why interferometer

Angular resolution higher

Better control of systematic errors

Lower cost of a large number of small antennas, with the same collecting area as a single large antenna

Telescope objectives

Aim for telescope

Single specific measurement

General purpose

expected lifetime

Does an existing facility already deliver this capability? why not use it?

Is an interferometer the right instrument for this objective

Requirements

derive technical requirements from the telescope objective

Many will be dependent on eachother

Multiple solutions can achieve the same objective

Similar sounding objectives give different performance

Optimisation to determine design choices will be needed

Cost likely to be deciding factor

Design choice areas

Angular scale of interest

Observing frequency

Sensitivity

Telescope location

Receptor type

Radio frequency interference

Configuration of the telescope

max and min scale size measured - larger objects are resolved out

Weighting of visabilities in the aperture plane determines point spread function so it is possible to tailor angular size sensitivity

Driver for spectral line measurements

Continuum observation driven by source spectral index

Choice of bandwidth

Bandwidth improves continuum senitivity and allows measurement of spectral index

Wider bandwidth allows more lines to be measured

Choice of channel bandwidth

Mostly unimportant for continuum

Ensure sufficient resolution for line observations

Blockage by secondary, supports etc.

illumination efficiecy

Surface errors, ruze formula

Higher frequency observations all higher bandwidths

Combining more than 30% fractional bandwidth into one image becomes problematic

System temperature

Dominated by 1st amplifier in receiver chain and losses before this amplifier

Generally cryogenicly cooled before this amplifier

Spillover due to idelobes of the telescope beam hitting arm ground

Depends upon illumination, blockage and zenith angle

Survey speed

Not relevant if aim is to observe individual, discrete objects

Large area,high frequency surveys are difficult

going to deeper depth quickly becomes prohibitive

Surface brightness sensitivity

If interested in imaging a large angular scale feature, need sensitivity on hort baselines

Need compact array of telescopes

Shadowing (looking into the back of another telescope at some angles) becomes an issue

Height above sea level important for precipitable water vapour content - ideally above the inversion layer

Radiofrequency interference from human habitation (a ground sheild or surrounding mountains can hepl

Ease of access to site

Infrastructure to run telescope

Local (expert) support for operations

Which regions of sky are accessable

Tropospheric stability

Ionospheric stability

Fraction of days when rain/wind/snow make observing impossible

The relative importance of the above issues depends critically on frequency

Dish antenna

Aperture array

station of several individual elements

electronics apply complex weights and sum signals

Form multiple beams on sky

Same as a dish bringing radiation to a focus

Particularly good at low frequency

Feed horn

Very well controlled beam, low sidelobes

Expensive to give large collecting area

Cylinders

Metal reflector form beam in 1D

Electronics can form beams in orthogonal direction

Issues

Many different choices of optics

offset/on axis; effects of blockage

prime/secondary focus (or tertiary)

Cassegrain/Gregorian/Dragone

Mounting arragement e.g. Alt/Az

elevation range

Pointing accuracy

Note that source positions on map depend upon visibility phase not antenna pointing

Antenna slew and settle rate

Reduce time lost driving to next field

Calibration cadence can enforce requirements

Alternatively use transient instrument

Antenna size

Larger area gives higher sensitivity

More expensive with bigger size (d^3)

Larger dishes require higher pointing accuracy

Surface accuracy - Ruze formula

At low frequency can use mesh rather than solid dishes

Polarisation

Receivers can be intrinsically sensitive to either linear or circular polarisation

Mesuring both orthogonal polarisations give 2 independent measurements of tokes

Requires a receiver chain or both and four correltors for each baseline

Astronomical radio signals are weak even for the strongest sources of interest

Many sources of man made RFI

Satalite down-links

Airport/military radar

Mobile phones

Television and radio transmitters

Sparkling from power transmission

RFI signals can be time variable and both narrow and relatively wideband

Waterfall plot of f/t used to identify and flag RFI

Interferometers help with RFI rejection

Only detect correlated rfi

Signals map to north or south pole

Self induced RFI

Sub-systems on the telescope itself can cause RFI

Any digital electronics (digitisers, correlator, computers

Drives and encoders

Power supplies

Screen individual components

Locate computers and correlator inside a screened room (Faraday cage)

Optical fibre rather than coax

Baseline length determines which scales are measured

The layout or configuration of the antennas in an interferometer determine sampling in the uv plane

To maximise filling of the uv-plane, ensure no redundant baselines (any structure in the configuration will be reflected in the psf)

Scattering antennas randomly over a wide observatory area has a large overhead in power, data transport, road etc. infrastructure

Reconfigurable antennas allows great flexibility but significant infrastructure overhead and observing time loss

Logarithmically space antennas along pirals from a centra core give a good compromise

Redundant baselines do offer additional possibilities for calibration

Correlator

Normally digital these days but analogue is possible

Niquist sample the time-stream data and digitise signal with n bits of resolution

Split into frequency channels

Apply delays, correlate and integrate over time

Size of correlator determined by number of baselines, number of frequency channels (not linea dependence

Different correlator modes (e.g. frequency zoom) increases flexibility and complexity

Compute requirements

For mapping, need to irst grid all the visibilities

If we desire to map entire FoV then can find number of pixels in image

Rarely do this in VLBI

Choose image size in pixels to be 2^n to allow use of fft

To avoid time averaging smearing, need correlator integration time less than maths

To avoid frequency smearing, need frequency channel bandwidth less than maths

Wide field mapping requires w correction (see anna's talk) - additional computational difficulties

Net effect of all of the above is that a continuum survey mapping experiment with a large Nant, small d and large Dmax can quickly become a Big Data challenge