Here are some streaming comments as I'm listening to results of the 2010 Decadal Survey. A shortlink to the report is here.
11:10 am EDT: Excellent placement of "what were the first luminous objects, and when did they form" second from the top of "Major questions to address this decade".
11:20 am EDT: Ooh, even better. "Cosmic Dawn" is the first of the 3 over-arching fields. Good fielding for EoR as a top science priority in the next decade!
11:23 am EDT: Now we're starting into descriptions of the panels. I was involved in several of the RMS (Radio, Millimeter, and Submillimeter) submissions, and am looking for HERA (the Hydrogen Epoch of Reionization Array). I'm also rooting for the Allen Telescope Array's Radio Sky Surveys Project. Finally, I'm hoping for some mention in the TEC (Technology development) program of prioritizing the development of shared solutions to digital signal processing hardware and libraries, which was the recommendation of a white paper I drafted with the help of the CASPER community.
11:45 am EDT: Mention of SKA as a priority for Radio Astronomy, unsurprisingly.
11:47 am EDT: Looks like Roger's wrapping up here. Turning to Q&A.
Meanwhile, I'm reading through the report. I see radio instrumentation is listed as a funding priority on ES-4, linking to 7-39. A good sign.
A painful line on 1-18: "U.S. participation in projects such as the Square Kilometer Array is possible only if there is either a significant increase in NSF-AST funding or continuing closure of additional unique and highly productive facilities." Ouch.
But on 2-12, my own sky map (well, with "permissions pending" for now). Now that's something!
12:01 pm EDT: An interesting question. "Why such a priority on habitable planets in the decadal review?" Sounds like the answer is that it's particularly primed to make big breakthroughs. I think I agree with that. I wasn't surprised that it was on there. There's some rumors going around that Kepler has found earth-like planets in earth-like orbits.
12:02 pm EDT: What about the surplus of post-docs relative to faculty positions? Answer: there are a wide range of careers and positions available to astronomers, so it's not unreasonable to have a larger number of post-doc positions where budding astronomers get training. I'm not sure that answer fully appreciates the scale of the problem.
Continuing with reading, on page 3-13, "The HERA program, a project that was highly ranked by the RMS-PPP and included by the committee in its list of compelling cases for a competed mid-scale program at NSF, provides a development pathway for the SKA-low facility. Progress on development of the SKA-mid pathfinder instruments, the Allen Telescope Array in the U.S., the MeerKAT in South Africa and the ASKAP in Australia, and in new instruments and new observing modes on the existing facilities ... will provide crucial insight into the optimal path towards a full SKA-mid." That's good to see mentioned. Sounds like it lays the groundwork for a strong future proposal to get funded. It's not a promise of funding, though. Not that such a promise was expected.
12:11 pm EDT: Second use of "tripwires" for projects. A very colorful phrase.
Interesting plot on 4-15: papers in all astronomy fields are increasing. Instrumentation papers seem to be low in number, but holding their own against other fields (same percentage contribution to total paper number).
On 5-14 for Data Reduction and Analysis Software: "Flexibility, openness, and platform independence, modularity, and public dissemination are essential to this effort. Focused investment in a series of small-scale initiatives for common tool development ... may be the most cost-effective approach, although there are undoubtedly synergies with the pipeline development needed for the large-scale projects." Sounds helpful for some of my projects like AIPY, SPEAD, and CASPER.
12:18 pm EDT: Comments on the SKA? Answer: SKA is the future, but the US can't pay for the construction on the proposed time scale (but slower might be ok). Technology development should be prioritized though. Low-frequency SKA, though, targets EoR, and we're interested in projects targeting that. Yes!
On 5-21 for Technology Development: "The committee received community input in the form of white papers on the funding needs for technology development in areas such as ... high speed, large N correlators. In these areas and others, researchers ... had come together to plan a coherent strategy for the decade. The OIR and RMS panels made a convincing case that the current level of ATI funding needs to be augmented in order to successfully pursue these highly-ranked technology development programs and roadmaps." Looks like my white paper fell on receptive ears.
12:28 pm EDT: Neil Tyson is closing down Q&A. He is one cool dude. I'm glad he was on the panel.
On 7-7 for Science Objectives for the Decade: "Find and explore the epoch of reionization using hydrogen line observations starting with the HERA telescopes that are already under construction." Wham. And in Table 7.1 on 7-32, Priority 2, Projects thought compelling: HERA.
On B-2 for Program Priorization, in Table B.1, I see both ATA and HERA. I'd say ATA didn't necessarily win big in the review, but at least they're there.
And finally, in appendix D-1: "Hydrogen Epoch of Reionization Array ... is a multi- stage project in radio astronomy to understand how hydrogen is ionized after the first stars start to shine. The first phase (HERA I) is under way and will demonstrate the feasibility of the technical approach. The second phase (HERA II) would serve as a pathfinder for an eventual world-wide effort in the following decade to construct a facility with a total collecting area of a square kilometer and the power to make detailed maps of this critical epoch in the history of the universe. Proceeding with HERA II should be subject to HERA I meeting stringent performance requirements in its ability to achieve system calibration and the removal of cosmic foreground emission." We've got our work cut out for us!
Showing posts with label signal processing. Show all posts
Showing posts with label signal processing. Show all posts
Friday, August 13, 2010
Friday, January 22, 2010
Where is GCC for FPGAs?
A lot of the digital signal processing that gets done in radio astronomy these days is done on Field Programmable Gate Arrays (FPGAs), and one of the projects I've been working on from the beginning in my research is developing open-source libraries for programming these chips. My part in this has generally been on the algorithmic/mathematical side: writing FFTs, filters, cross-correlation engines, etc. Another key aspect of this work, though, is a toolflow that allows people to design systems at a high level with parameterized algorithmic cores, and to turn that design into the wiring instructions that tell the FPGA how to implement the system.
We currently use a design entry system based on Simulink running on Matlab, and while it is an extremely powerful environment, we've also found it to be limiting, frustrating, and hard to maintain designs in. In October, I volunteered at an international workshop on astronomy signal processing to explore alternatives to this environment. My current favorite is MyHDL, which uses Python to generate lower-level code in Verilog or VHDL, and I may start looking more deeply into porting a design to use MyHDL.
Something that is bothering me, though, is that however much we work on porting our toolflow open-source equivalents, there is currently no open-source compiler for FPGAs. The state of affairs in FPGA-land is something like PCs in the '70s, when every personal computer had its own specialized compiler. For PCs, the problem was solved by GCC (the Gnu Compiler Collection), which became the default open-source solution for compiling most languages to target the many CPU architectures that exist in the world today.
I'm keeping my eye on gEDA, and notably Icarus, which seems to be a free synthesis tool (synthesis, mapping, and routing are the 3 main stages of compiling for an FPGA). Perhaps mapping and routing can never be open-source, since they tend to be very chip-specific. But here's hoping...
We currently use a design entry system based on Simulink running on Matlab, and while it is an extremely powerful environment, we've also found it to be limiting, frustrating, and hard to maintain designs in. In October, I volunteered at an international workshop on astronomy signal processing to explore alternatives to this environment. My current favorite is MyHDL, which uses Python to generate lower-level code in Verilog or VHDL, and I may start looking more deeply into porting a design to use MyHDL.
Something that is bothering me, though, is that however much we work on porting our toolflow open-source equivalents, there is currently no open-source compiler for FPGAs. The state of affairs in FPGA-land is something like PCs in the '70s, when every personal computer had its own specialized compiler. For PCs, the problem was solved by GCC (the Gnu Compiler Collection), which became the default open-source solution for compiling most languages to target the many CPU architectures that exist in the world today.
I'm keeping my eye on gEDA, and notably Icarus, which seems to be a free synthesis tool (synthesis, mapping, and routing are the 3 main stages of compiling for an FPGA). Perhaps mapping and routing can never be open-source, since they tend to be very chip-specific. But here's hoping...
Monday, November 9, 2009
The Need for SPEAD
I've been absent for a good while now as a result of participating in a (successful) deployment of our PAPER experiment in South Africa. The Karoo desert in SA, where we were stationed, was very reminiscent of Rangely, CO where I grew up, except for the occasional baboon or kudu in the road. Though it came at a price of a lot of work piled up for me when I got back, and an awfully long time away from J, the isolation from all but our experiment helped ferment some ideas I'd been having about migrating the AIPY toolkit I've been developing to use a streaming data format that would avoid unnecessary disk accesses, would allow AIPY to be integrated directly with the correlators developed by our CASPER project, and would help our experiment develop a real-time analysis pipeline for compensating for ionospheric distortion in our data.
After chatting with a lot of guys working on the Karoo Array Telescope in Cape Town, we came up with a concrete protocol build on something already being used for CASPER correlator output. I just got done writing my first grant proposal to the NSF, funding a graduate student to work on this protocol--the Streaming Protocol for Exchanging Astronomical Data (SPEAD, pronounced "speed"). The process of writing a grant myself was a learning process, and helped me understand where a lot of the questions I got asked by my previous advisors were coming from.
A lesson I got to take away from SA was this: the reason we were in SA (as opposed to Australia) for PAPER was because we had been working with the KAT team, sharing correlator development. The reason we were working with the KAT team was because CASPER and KAT started up a collaboration a few years before. And that collaboration was started up because Dan Werthimer went down to visit SA some years ago to help advise them in a review of the design of their telescope electronics. Dan was invited there because he struck up a fast friendship with Alan Langman (the KAT director) at an earlier conference. The moral of this chain of causes and effects being that sometimes large projects go in new directions because of personal friendships, and sometimes those friendships end up making the difference in the success of a project.
Wednesday, August 26, 2009
Graph-SLAM
After a trying, but ultimately successful month spent extracting the family from Puerto Rico and re-embedding us in Berkeley, I'm just starting to get back on top of things enough to think about posting...
I had lunch yesterday with a good friend of mine, Pierre, who co-founded a company that specializes in sensory and mapping systems such as those that are used to create Google's "Street View". I was impressed to learn about their system for combining data from GPS, LIDAR, car odometers, and IMUs to create a consistent picture of how a vehicle is located and oriented in space as a function of time. They've spent a lot of time calibrating their systems, and use some sophisticated MCMC post-processing methods for deriving the actual trajectory of a vehicle.
Although the antennas in the PAPER array (that's the low-frequency interferometer I'm working on), are much less mobile than a car, there was considerable overlap between the problem Pierre has been working to solve and the calibration problem I am facing the requires positioning antennas and celesital sources as a function of time in the face of ionospheric distortion, variable gains, etc. Pierre pointed me to Graph-SLAM as a formal description of the problem that we are trying to solve, and suggested that Kalman Filtering with RTS Smoothing was a powerful technique for converging to the optimal solution (with covariance information) in linear time.
I had lunch yesterday with a good friend of mine, Pierre, who co-founded a company that specializes in sensory and mapping systems such as those that are used to create Google's "Street View". I was impressed to learn about their system for combining data from GPS, LIDAR, car odometers, and IMUs to create a consistent picture of how a vehicle is located and oriented in space as a function of time. They've spent a lot of time calibrating their systems, and use some sophisticated MCMC post-processing methods for deriving the actual trajectory of a vehicle.
Although the antennas in the PAPER array (that's the low-frequency interferometer I'm working on), are much less mobile than a car, there was considerable overlap between the problem Pierre has been working to solve and the calibration problem I am facing the requires positioning antennas and celesital sources as a function of time in the face of ionospheric distortion, variable gains, etc. Pierre pointed me to Graph-SLAM as a formal description of the problem that we are trying to solve, and suggested that Kalman Filtering with RTS Smoothing was a powerful technique for converging to the optimal solution (with covariance information) in linear time.
Tuesday, May 5, 2009
Compressed Sensing and Wiener Filtering
Today I'm trying to expand my understanding of how we can best remove contaminant signals from the data we take with the Precision Array for Probing the Epoch of Reionization (PAPER). There is a specific problem I want to make sure we can solve for PAPER. Foregrounds to our signal, particularly synchrotron radiation, are expected to be very smooth with frequency. The idea put forth by the MWA and LOFAR groups is that by observing the same spatial harmonics at multiple frequencies, we should be able to remove such smooth components to suppress them relative to the cosmic reionization signal we are looking for. However, generating overlapping coverage of spatial harmonics as a function of frequency is expensive. My intuition is that since foregrounds do not have a spatial structure that changes dramatically with frequency, we shouldn't need to sample a given spatial harmonic very finely in frequency to get the suppression we want. This would allow us to spread our antennas out a little more and get measurements of the sky at a variety of spatial modes.
In many ways, our problem is analogous to what was done with the Cosmic Microwave Background (CMB). For foreground removal in CMB work, Tegmark and Efstathiou (1996) begin with an assumption that foregrounds can be described as the product of a spatial term and a spectral frequency term. This allows them to construct Wiener filters that use the internal degrees of freedom of their data, together with a model of their foreground and a weighting factor based on the noisiness of their data, to construct a filter for removing that foreground. For the most part, this is standard Wiener filtering, except they have to be careful about what they do to their power spectrum, so they apply a normalization factor to correct for a deficiency in Wiener filters. Tegmark (1998) goes on to generalize this technique for foregrounds that vary slowly with frequency. I'm in the process of wading through these papers, but they seem to be directly applicable to what we are doing, and seem to confirm my suspicions that synchrotron emission should be well-enough behaved to require only sparse frequency coverage of a wavemode in order to be suppressed.
Another tactic that I am investigating is that of compressed sensing which I was alerted to in talks by Scaife and Schwardt at the SKA Imaging Workshop in Socorro this last April. The landmark paper on this principle seems to be Donoho (2006), where it is shown that the compressibility of a signal (being sparse for some choice of coordinates) is a sufficient regularization criterion to faithfully reconstruct signals using a small number of samples. In a way, this technique has an element of Occam's Razor in it--it tries to find a solution, in some optimal basis, that needs the fewest non-zero numbers to agree with the measured data. At least, that's my take on it without having finished the paper.
The relevance of compressed sensing to image deconvolution is explored in Wiaux et al (2009), and it seems to be powerful. I'm excited by this deconvolution approach because it meshes well with the intuitive approach I've been taking to deconvolution, which was to use wavelets and a Markov Chain Monte Carlo optimizer to find the model with the fewest number of components that reproduces our data to within the noise. Compressed sensing seems to be exactly this idea, but is agnostic about the basis chosen, instead of mandating one like wavelets. Anyway, this technique may also be relevant to our foreground removal problem because we might be able to use it to construct the minimal foreground model implied by our data. For synchrotron emission, which should have smoothly varying spatial structure with frequency, I envision that this could construct a maximally smooth model that would allow us to use sparse frequency coverage to remove the foreground emission to the extent that it is possible to do so.
In many ways, our problem is analogous to what was done with the Cosmic Microwave Background (CMB). For foreground removal in CMB work, Tegmark and Efstathiou (1996) begin with an assumption that foregrounds can be described as the product of a spatial term and a spectral frequency term. This allows them to construct Wiener filters that use the internal degrees of freedom of their data, together with a model of their foreground and a weighting factor based on the noisiness of their data, to construct a filter for removing that foreground. For the most part, this is standard Wiener filtering, except they have to be careful about what they do to their power spectrum, so they apply a normalization factor to correct for a deficiency in Wiener filters. Tegmark (1998) goes on to generalize this technique for foregrounds that vary slowly with frequency. I'm in the process of wading through these papers, but they seem to be directly applicable to what we are doing, and seem to confirm my suspicions that synchrotron emission should be well-enough behaved to require only sparse frequency coverage of a wavemode in order to be suppressed.
Another tactic that I am investigating is that of compressed sensing which I was alerted to in talks by Scaife and Schwardt at the SKA Imaging Workshop in Socorro this last April. The landmark paper on this principle seems to be Donoho (2006), where it is shown that the compressibility of a signal (being sparse for some choice of coordinates) is a sufficient regularization criterion to faithfully reconstruct signals using a small number of samples. In a way, this technique has an element of Occam's Razor in it--it tries to find a solution, in some optimal basis, that needs the fewest non-zero numbers to agree with the measured data. At least, that's my take on it without having finished the paper.
The relevance of compressed sensing to image deconvolution is explored in Wiaux et al (2009), and it seems to be powerful. I'm excited by this deconvolution approach because it meshes well with the intuitive approach I've been taking to deconvolution, which was to use wavelets and a Markov Chain Monte Carlo optimizer to find the model with the fewest number of components that reproduces our data to within the noise. Compressed sensing seems to be exactly this idea, but is agnostic about the basis chosen, instead of mandating one like wavelets. Anyway, this technique may also be relevant to our foreground removal problem because we might be able to use it to construct the minimal foreground model implied by our data. For synchrotron emission, which should have smoothly varying spatial structure with frequency, I envision that this could construct a maximally smooth model that would allow us to use sparse frequency coverage to remove the foreground emission to the extent that it is possible to do so.
Subscribe to:
Posts (Atom)