Nebula: goals

High-level goals

A science experiment should have a conclusion, positive or negative. After 18 years, SETI@home does not yet have one. My goal in developing Nebula is to finish what we started in 1998 - to reach a point where we can write a scientific "results" paper that

In addition to providing a conclusion to SETI@home, this paper would expose the SETI@home methods and algorithms to the world, so that people outside our group (other SETI researchers, signal-processing experts, etc.) can critique them and suggest improvements. Perhaps our data contains an ET signal but our current algorithms can't find it.

Publishing this paper won't mark the end of SETI@home. It may continue indefinitely. However, it will probably undergo a major transformation soon - new telescopes, new frequency bands, new front-end algorithms - that will effectively start a new experiment. We need to finish the original experiment.

RFI strategy

SETI@home is a wonderfully sensitive detector of possible ET signals. Unfortunately, it's also a great detector of RFI that resembles ET signals.

Before Nebula, our RFI removal algorithms removed perhaps 95% of RFI. When we looked at our top-scoring multiplets and examined their "waterfall plots" - the graph of signals in time/frequency space - we could see that they were obvious RFI, part of the 5% that our algorithms didn't remove. The proposed strategy for dealing with this was:

We discussed training volunteers to do this, in the style of projects like Stardust@home.

However, even with the help of volunteers, this strategy was never feasible. A multiplet can include hundreds of signals, and it would take hours to examine them all. And there are millions of multiplets. If one of them were an ET signal, it could be in the middle or the bottom of the list, and we'd find it only after manually examing everything above it, which could take decades.

Nebula takes a different approach to RFI. Its goal is to iteratively develop RFI algorithms that remove almost all - say, 99.999% - of the RFI (hopefully without removing ET signals as well). An ET signal would then be fairly close to the top of the list of top-scoring multiplets.

In this approach, we still have to manually examine the waterfall plots of top-scoring multiplets, looking for obvious RFI. When we find it, rather than flagging those particular frequencies, we return to our algorithms and figure out how to change or extend them to remove not just that particular RFI, but everything like it.

Specific goals

Thus, I want Nebula to speed up the back-end tasks - RFI removal and persistent signal finding - to a point where the entire process, started from scratch, finishes in a day or two. This will, as described above, allow the algorithms for these tasks to be improved iteratively. Nebula should let our scientists (e.g. Eric Korpela and Dan Werthimer) do the following:

  1. Do the back-end processing in its entirety.
  2. Examine the top-scoring multiplets and pixels. If their constituent signals look like RFI, figure out how to modify or extend the RFI algorithm to remove this type of RFI.
  3. Repeat 1 and 2 until a significant fraction of the top-scoring multiplets are not obviously RFI.
  4. Do re-observations.
Steps 1 and 2 may need to be repeated dozens of times; this is infeasible if they take months to complete.

Other goals of Nebula include:

Next: Nebula architecture

©2019 University of California
SETI@home and Astropulse are funded by grants from the National Science Foundation, NASA, and donations from SETI@home volunteers. AstroPulse is funded in part by the NSF through grant AST-0307956.