My focus these days is on the processing of birdie signals - are they getting flagged as RFI? Are we finding the multiplets we'd expect?

Toward this end I added beam number to the data shown about birdie signals on the web. To my surprise, almost all were from beam 0! I took a closer look: about 10,000 signals in beam 0, 50 in beam 1, ... 4 in beam 6. This is wrong- all beams should produce roughly the same number.

I looked at the code. There's a variable keeping track of the time of the last signal generated - new ones can only generated a certain amount after this. I had failed to reset it at the start of each beam. I fixed this and now we're getting on the order of 10,000 signals per beam, like we should.

I reran the pipeline, the results of which are now online. It's immediately clear that a large fraction of birdie signals are being flagged as drifting RFI. Apparently there's a problem with the drifting algorithm.

Other issues: way too many pulses and triplets (non-birdie) are flagged as multibeam RFI. And multiplet finding for birdies seems wrong. For each birdie, it finds lots of multiplets with a few signals each. Ideally it would find a single multiplets with a lot of signals (and a high score).

Plugging away. This will converge eventually.