Joined: 13 Feb 99
The drifting RFI algorithm I added recently (an adaptation of the one used for decades by SERENDIP) was effective for removing drifting RFI. However, it also removed about 90% of all the other signals.
Recall that the algorithm works as follows: for each signal, we look at 6 radial sectors in time/freq space, going forward in time with different slopes (corresponding to different ranges of drift rate). If any sector has a number of signals that exceeds a threshold, its contents are marked as RFI.
Using the RFI browsing tools I described in the last post, I explored several examples of this, and found that there were two problems:
1) Spikes often occur in "clusters": 5-10 spikes at almost the same time and freq. If there was a signal somewhat before this, at a close freq, the sector containing the cluster would exceed threshold and all the spikes would (erroneously) get thrown out.
2) Where there is narrow-band RFI (drifting or not), signals within a few 1000 Hz were getting thrown out because one of the oblique sectors from that point would include some of the RFI.
Eric Korpela and I spent some time discussing these problems. In the end, the following seemed to work in the cases I looked at:
- For 1), identify clusters (+/- 1Hz, +/1 1 sec) and treat them as 1 signal.
- For 2), change the RFI algorithm parameters to:
max drift rate: 5
With these changes, the drifting algorithm still removes the target RFI, and the fraction of spurious removals is down to < 1%.
Joined: 18 Apr 09
Thanks for the updates. Really appreciating that we can actually see how things are moving forward (and even kind of touch and touch and try out...)
Joined: 27 Feb 17
thank you very much!
©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.