The goal of SETI@home is to find ET. But if we don't, we want to make a statement of the form "There are no radio beacons in the Arecibo sky, in the S@h frequency range, with power greater than X". Of course, we want to make X as low as possible; it represents the sensitivity of our search.

However, it's trickier than that. A radio beacon has various properties: its bandwidth, whether it's pulsed, how its observed frequency varies over time, and so on. Our sensitivity depends on all these factors. To complete our study, we'll need to consider various combinations of factors, and estimate X for each one.

Our birdie mechanism lets us do this: for a given set of parameters we can create birdies with varying power, run them through the pipeline together with the real data, and see which birdies we detect.

Right now we're focused on how the frequency varies over time. If ET is transmitting at a constant frequency, we'll detect it at a frequency that varies over time due to the acceleration of the transmitter. So when we generate signals for such "non-barycentric" birdies, we model this variation and add it to the detection frequency.

If the transmitter is orbiting a star, its signal will be shifted by the orbital acceleration.
If in addition it's on the surface of a planet, or is orbiting a planet, it will be shifted according to the rotation or orbit.

Until recently we modeled only stellar orbit. Eric Korpela recently extended this to model transmitters on or orbiting a planet. We can now model ET beacons with a range of stellar and planetary parameters, including planets both in and not in the star's habitable zone.

This is pretty cool. We can now realistically model many types of ET signals, and measure our sensitivity to each one. SETI@home is unique in having this capability, which is especially useful because of the wide range of signal types that SETI@home can detect.

Other progress:

I added the ability to have waterfall plots show beam number rather than signal type. This is useful for identifying certain types of RFI. Click the "Plot by beam" button in a waterfall plot.

We widened the frequency window for barycentric multiplets from 10 Hz to 125 Hz. There were various sources of frequency uncertainty adding up to the latter.

We added an exclusion zone of +/- 1220 Hz around 1420 Hz. Signals in this range are typically strong DC bin signals or harmonics.

At the end of each Nebula pipeline run, we now create and archive a "snapshot" consisting of the crucial output files (multiplets and pixel scores). This lets of study the effects of changes in algorithms and parameters.

From the beginning of your message it would appear that you're currently predicting a 'no ET' scenario (at the current levels of sensitivity)
However, are there any pixels that are consistently showing a statistically significant excess of possible ET signals?

---

We're still in the stage of working on the algorithms,
which still have room for improvement.
At this point, when we do a "run", we process only 100K out of 16M pixels,
and they're randomly chosen so different every time.

Currently we're focusing on birdies; i.e., improving the algorithms to find faint birdies.
If we can't find faint birdies, we can't find faint ET.
When we think the algorithms are "good enough" we'll do a full scoring run.

-- David

Thanks for keeping us updated on your progress with Nebula! It seems to be a difficult process to try to remove RFI to weed out any false positive signals without losing any potential real signals.