Hugh Garsden's Work Page

I work at the Harvard-Smithsonian Center for Astrophysics within the LEDA project, on several astrophysics projects using data from the OVRO-LWA radio telescope in California. My work includes:
  • 21cm cosmology at Cosmic Dawn/EoR
    • Observations of the global HI signal
    • Observations and power spectra
  • Pulsar observations
  • Radio interferometry
  • Data processing
  • Data analysis
  • Telescope management
When I arrived at CfA I supported the establishment of the OVRO-LWA telescope, working on software pipelines, calibration and imaging, data quality checks, data storage, cataloging, RFI detection, monitoring and control. I ended up taking over most of this for LEDA, as my time went on. The telescope is generating data that is used for research at CfA and CalTech.

Here is a "short" summary of what I've been up to recently:

I'm currently working on generating power spectra of the 21cm HI signal at Cosmic Dawn, integrating 4 hours of telescope data over a 10MHz bandwidth at 48MHz. 
I'm also simulating power spectra of telescope noise to get an idea of the sensitivity we can acheive if we integrate lots of observations. The telescope generates a huge amount of data, so I built an in-house high-performance power spectrum generation pipeline to do the processing (C++ program) in a reasonable time, including some GPU code for noise simulations.

Mike Eastwood at CalTech has published power spectra generated using spherical harmonic analysis. I am using the delay-spectrum method which is a common method and used by other groups from other telescopes like MWA, LOFAR, PAPER, HERA. Our results will be coming out in a paper soon (2020). Like most others, our power spectra are not high-quality enough to get a 21cm signal detection, as there are errors in the telescope data and analysis that we all have to work on beating down. However, the simulations of telescope sensitivity are quite interesting, and I'll leave that for our publication!

I also work on 21cm global signal cosmology. The LEDA group is one of a few around the world following up the global signal detection reported by Bowman.
We have gathered several hundred hours of sky spectra and I'm analysing them
using a pipeline developed by Danny Price. I'm also generating RFI detection algorithms. Once we have a set of clean data we run it through a Bayesian/MCMC analysis to see if we have a detection. There will be a couple of papers coming out of this - initially we want to report on the observations, data quality, what we've got so far and where we are headed.

I have also made some observations of pulsars through the Sun's corona. LEDA had a project to measure the magnetic field close to the sun, but we were beaten to it by some of our collaborators. It did spur the development of a new software pipeline system combining CPUs and GPUs -> "Bifrost". A paper was published on that (in my ref list below) and it has a life of its own now.

The continuing ride:

Apart from just trying to get better data quality, calibration, and flagging, work on power spectra at LEDA is going to move into simulations (IMHO) - understanding the issues with the OVRO-LWA telescope, how they are affecting power spectrum generation, how they might be fixed, what power spectra we should be able to achieve, what sensitivity level, and what the power spectra will tell us. There is a hint of that in the upcoming paper. Power spectrum data from several telescopes have been presented in the literature, but a real detection will be made with new telescopes like HERA and the SKA. There are lots of simulations tools being  generated for these telescopes that LEDA could use, I've investigated some, but haven't had the time for anything in detail.

Previous position

From 2012-2014 I was at CEA, south of Paris, in the CosmoStat Laboratory. I was employed by the Université-Paris-Diderot. I implemented compressed sensing for radio astronomy imaging in the LOFAR Radio Telescope, the first (I believe) implementation of compressed sensing in a working telescope imager.

I was a member of the LOFAR Slow Transients Software Team and the LOFAR Source Finder Working Group. Finding "slow" transients involves searching radio images for astronomical sources and comparing them against known sources. This has to be done in an automated way using software, due to the huge number of images generated. I worked mainly on source finding, software and testing.


I worked on gravitational microlensing and built the first microlensing software simulator that could incorporate 1 billion microlenses. Using this I published several papers studying lensing in this new region of parameter space.

Previous lives

Prior to getting my PhD I worked as a software engineer in industry and at universities in research groups in physics, computing, speech synthesis, and biomedicine.


  • PhD, University of Sydney. Gravitational lensing, numerical computing, and supercomputer simulations.
  • B.Sc. (Hons I), University of Adelaide and University of Queensland. Physics, Maths, Computing.

Research Interests

  • 21cm cosmology
  • Numerical computing, algorithms, parallel computing, GPUs, deep learning, tensor flow
  • Cosmology
  • Simulations in these fields
  • Radio astronomy, interferometry, algorithms
  • Pulsars
  • FRBs
  • Transients
  • Multi-messenger astronomy
  • Gravitational lensing
  • Quasars
  • Supercomputing, Big Data
  • Signal and image processing


A representative sample:


Please view my list on SAO/NASA ADS.

plus 14 papers from my previous lives.


I'm on Research Gate, LinkedIn, personal stuff goes on Facebook. I maintain a blog for keeping my research notes; these are not meant as a source of information for the public. It serves as my notebook, so I can keep track of things.

Email: hugh.garsden at