Precision Simulations of Cosmic Dawn Experiments Adam Lanman Brown - - PowerPoint PPT Presentation

Precision Simulations

• f Cosmic Dawn Experiments

Adam Lanman – Brown University Science at Low Frequencies III December 8th, 2016

EoR vs. Foregrounds

• Foregrounds (radio galaxies, SNRs, galactic synchrotron, thermal

bremsstrahlung, etc.) are typically 105 x brighter than the expected EoR signal.

Coordinates

• Visibility coordinates correspond with spatial

coordinates:

– u is already in Fourier space. – Frequency corresponds with distance, so its

Fourier dual (η) corresponds with k||.

Comoving distance

Foreground Avoidance

• Smooth-spectrum foregrounds

will go to low k||, while the complicated spectrum of the EoR will appear at high k|

• Instrument chromaticity will throw

some power to higher k|| into a characteristic wedge shape.

MWA Phase II PAPER128 HERA19

• Murchison Widefield Array
• Precision Array for Probing

the Epoch of Reionization

• Hydrogen Epoch of

Reionization Array

MWA Data Analysis

• Preprocessing – Raw correlator output is reformatted and

flagged for RFI contamination.

• Calibration, forward-modeling, and imaging are all done in FHD.
• eppsilon does power spectrum estimation

Fast Holographic Deconvolution

(Convolution) (Gridding) Holographic mapping function

Fast Holographic Deconvolution

• The baseline positions and a saved antenna beam model are used to construct a point spread

function for the array.

– Beam models are made using electromagnetic simulations.

• The beam model is used to construct and then grid model visibilities based on a source catalog.
• The model visibilities are used for calibrating the raw visibilities from the data file.

Weights Cube

Beam Model Beam Model Beam Model

FHD as instrument simulation

• All that's needed to generate the model visibilities is a set of baseline

coordinates, a beam model, and a source catalog.

• EoR signal, diffuse emission, and noise can also be included.

Weights Cube

Beam Model Beam Model

Simulating EoR

• FHD may be used to

simulate an EoR signal.

• Gaussian Sky

– Generate random

brightness fluctuations

• n a sky with a given

power spectrum.

• Bubble Simulation

– Tile a cube from a

21cmFAST simulation.

εppsilon results of an EoR simulation using the MWA128 configuration. Red is the input EoR signal, and blue/black are the signal recovered from the instrument.

PAPER Pipeline

• A quadratic estimator is used to calculate the power spectrum.
• This can include inverse covariance weighting and windowing.
• Errors are estimated by bootstrapping

– Select random subsamples of the visibilities and repeat PS estimation.

Figure adapted from Ali, et al. (2015)

Simulations

• Tests with a single 100m east-west

baseline.

– PAPER beams vs. MWA beams. – Adjusting simulation and phasing parameters.

• 9 hours worth of data from PAPER128

– Julian Dates 2456535.333 to 2456535.708 – RA = 0 crosses the zenith about halfway

through.

• 9 hours worth of data from HERA19

PAPER128 Results

Logarithmic waterfall plot of a single baseline. 10,000 GLEAM catalog sources simulated, with no added noise or EoR signal. The increased power later on is due to Fornax A entering the beam.

PAPER128 Results

(Horizon line)

HERA19 Results

“Unfortunate Lines”

• Delay-transformed waterfall

plots of FHD-simulated data show excess power at high delays.

• This seems to be due to the

resolution of the gridding kernel.

The top half shows data simulated with uv resolution of 0.5 λ. The bottom has resolution 0.1λ. Both are for HERA19 on an 81 m baseline with 300 sources, same times. (Lines are not present in actual data)

File Edges

• FHD only calculates

the primary beam

• nce per snapshot.

Measured brightness

• f sources changes
• ver time as sources

move through the beam.

• Simulated visibilities

show distinct “jumps” between files as a result of this.

• This is especially
• bvious with only one

source.

Single 100m east-west baseline with a single 100Jy source approaching the zenith.

Rounding Errors

• Snapshots are

phased to zenith at a given time step.

– Discontinuities appear

at these times.

• This came up when

phasing data with the pyuvdata package.

• Investigating this

error, we uncovered a bug in how FHD handles small uvw coordinates.

300 sources, simulated for a 100m east-west baseline and MWA beams.

Summary

• FHD's forward-modeling feature may be

used for standalone instrument simulation.

• Simulated data from PAPER128 and

HERA19 have been processed using the PAPER power spectrum pipeline.

• Generating raw visibilities with FHD has

revealed processing artifacts that must be understood.