Interface with pyuvdata/pyuvsim

[steven-murray]

There is some support currently for setting up by reading from a pyuvdata object. It would be nice to fully test this, and round it out by enabling reading pyuvsim config files.

[steven-murray]

For pyuvdata, it would be useful to use the pyuvdata object itself to provide the UV coverage, then use the standard mechanisms to get the sensitivity.

[steven-murray]

Might need to use some kind of aggressive BDA to make it more memory efficient, @jpober suggests.

[steven-murray]

To be honest, when I go and try to implement this, I find that I don't think there's much use in accepting eg. pyuvsim configs. The intent of those configs is just so much different from 21cmSense. I think instead we just try to maintain a good configuration for 21cmSense itself.

Thoughts, @mkolopanis ?

[steven-murray]

Note that the current code has support at least for creating the Observatory from uvdata objects (i.e antenna positions), which is probably the biggest thing that you'd want.

[mkolopanis]

The only other thing that would come to mind to me is extracting the total integration time from a pyuvsim config. I don't really think it's a big deal though. Seems reasonable to just leave it at accepting a uvdata object.

[jpober]

Yeah, I don't see a ton of value in accepting a pyuvsim config file if it's really totally different. Accepting a UVData object is still valuable though, as it's nice to ask "I have these observations actually in hand, what's their power spectrum sensitivity?"

[steven-murray]

At the moment, we get the antenna positions and latitude from the UVData object. Is there anything else we should be pulling? Would it be useful at all to set the total observation time from a UVData object? Seems to me like typically this would be a bad idea since a single object will have a very limited number of times in it.

I guess setting the frequency to the middle of the object's freq_array might be useful? There's Trcv but I don't think uvdata objects have that info. All other parameters that I can think of are not specified in a UVData object.

[jpober]

For one object it's not that useful, I agree. I imagined taking an entire lst-binned observation and building up a uv coverage from that. My comment at the top of this thread is that you perhaps want to do BDA first, so that you're not dealing with a sample every couple of seconds.