add demo_QE notebook (and use NamedTuple for exps)

[cmbant]

If you want a basic QE demo, something like this seems to work (and shows as expected that skewness of QU is much lower than T recon)

[cmbant]

Might want to tidy up for consistent use of gradient responses etc

[carronj]

Thanks. I believe you adding the 'name' to get_config will break some codes. But can include something similar

[cmbant]

Yes, sorry didn't test with anything else. Just wanted some way to play with expt config in notebook without having to modify the list of recognised definitions in the source.

[cmbant]

Btw, weird stuff happens if you change LDres, HDres, eg.. kappa looks crazy if I set HDres = 9

[carronj]

It makes little sense to LDres (~the data resolution) higher than HDres (~the sky resolution), but otherwise (e.g. both at 9) should work?

[cmbant]

No, that looks crazy too. Maybe I misunderstood what they are - from the get_maps_lib doc string I thought HDres was used to change the physical patch size, and LDres the number of pixels in the patch.

[carronj]

Thanks. with HD 9 it looks like the response at L close to the high-L limit 2 * lmax_cmb is so tiny and has(I guess) numerical noise that gets blown up by inverting this to normalize phi. These super-high Ls are useless modes anyways, so simplest is to truncate the ellmax of lib_qlm to something reasonable (like lmax_cmb)

[cmbant]

I was trying to just change the patch size to something smaller, with resolution, lmax etc all the same. It seems there should be three numbers, the data res, sky res, and pixel number, what is the intended way to do this?

[carronj]

What you wrote is correct, e.g. (9 10 ) gives the same as (10, 11) but on a box with sides smaller by a factor of two. (to avoid the kappa real-space you had, just better not use the lensing L ~ 6000 in remapping to real space)
(The sky resolution is fixed to 0.74 amin using the sim libs in this way, but one can define any othes adapting the sim lib definitions )