Need better X-WF-NT-NF designation

[o-smirnov]

The filter designation could be improved. NT-NF refers to the number of overlapping timeslots and frequency channels, correct? If so, this number will change depending on the sampling of the hi-res MS, and also mean different things (in terms of window width) depending on the lo-re sampling. Wouldn't it be better to indicate the degree of overlap in units of averaging interval?

To make sure I understand the current designation correctly, please confirm that the following is true:

• most experiments in the paper refer to a 1.5s, 125 kHz hi-res MS, resampled to 150s, 6.25 MHz (100 timeslots, 50 channels).
• A Bl-WF-100x50 overlap filter means an overlap of 100, 25 on each side. So the total window width is 100+100+100=300 timeslots and 25+50+25=100 frequencies, right?

Except then in the section on "theoretical noise estimate and simulated noise", you talk of 1min30s sampling (that's 90 seconds/60 timeslots, are you sure? Why not 150s/100 timeslots as in the other section?) and 2min60s and 3min45s overlap (very confusing time units!), and 12500 kHz sampling, and 3125 and 6250 kHz overlaps. That's 60 timeslot averaging, 120 timeslot overlap (2min60s?), then 150 timeslot overlap (3min45s). And 100 channel averaging, 25 and 50 channel overlaps. Please check that these numbers are correct, because I find the choice of intervals inconsistent with the rest of the paper.

I suggest:

• Using only seconds for all time intervals (i.e. 150s not 2m30s)
• Designating overlap in terms of effective window size, in units of the resampling interval. E.g. "Bl-WF-100x25" with 100 timeslot, 50 channel resampling should become "Bl-WF-3x2", since we're computing the window over x3 the number of timeslots and x2 the number of channels, compared to a non-overlapping filter.

[atemkeng]

All questions are corrects except the following:

Except then in the section on "theoretical noise estimate and simulated noise", you talk of 1min30s sampling (that's 90 seconds/60 timeslots, are you sure? Why not 150s/100 timeslots as in the other section?) and 2min60s and 3min45s overlap (very confusing time units!), and 12500 kHz sampling, and 3125 and 6250 kHz overlaps. That's 60 timeslot averaging, 120 timeslot overlap (2min60s?), then 150 timeslot overlap (3min45s). And 100 channel averaging, 25 and 50 channel overlaps. Please check that these numbers are correct, because I find the choice of intervals inconsistent with the rest of the paper.

When I was writing this part, I think of using the same resampling in time and frequency after I have wrote this part and complete forgot to update. The plots are under the same re-sampling (150s, 6.25Mhz) without overlap. sinc-100x25 total number of timeslots is 100+100+100 and channels 25+50+25 as you said earlier. If you see the table of comparison you will noticed that the resampling were understand the same conditions. Thank you

[o-smirnov]

OK got it. I'm going to simplify the designations as suggested above, i.e. sinc-3x2 instead of sinc-100x25. Also makes the math easier to formulate. Sorry, I'm omitting your definitions for the paper, I find them too formalistic and overly complicated.