bug in coarse channel frequency calculation

[lacker]

I think there's a bug in the code that calculates the center frequencies of the coarse channels - https://github.com/UCBerkeleySETI/hpguppi_proc/blob/mark_r/src/hpguppi_coherent_bf_thread.c#L384

Basically this expression:

((n_chan_per_node-1)/2)

since it's an integer expression, it's doing integer arithmetic, which means truncating the result of a division. I think the intention of this code is to calculate the center of the channel and in that case it should be doing a floating point division.

There's a similar issue in the upchannelizing version of the code and perhaps elsewhere.

[david-macmahon]

Yes, I think nchan_chan_per_node should be cast to a double like this:

((((double)nchan_per_node)-1) / 2)

or change 1 to 1.0 like this:

((nchan_per_node-1.0)/2)

[mruzinda]

Thanks for pointing this out. I'll make that change in the next commit.